agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v2 1/3] Add CREATE OR REPLACE MATERIALIZED VIEW
197+ messages / 2 participants
[nested] [flat]

* [PATCH v2 1/3] Add CREATE OR REPLACE MATERIALIZED VIEW
@ 2024-05-21 16:35 Erik Wienhold <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Erik Wienhold @ 2024-05-21 16:35 UTC (permalink / raw)

---
 .../sgml/ref/create_materialized_view.sgml    |  15 +-
 src/backend/commands/createas.c               | 207 ++++++++++++++----
 src/backend/commands/tablecmds.c              |   8 +-
 src/backend/commands/view.c                   | 106 ++++++---
 src/backend/parser/gram.y                     |  15 ++
 src/bin/psql/tab-complete.c                   |  15 +-
 src/include/commands/view.h                   |   3 +
 src/include/nodes/parsenodes.h                |   2 +-
 src/include/nodes/primnodes.h                 |   1 +
 src/test/regress/expected/matview.out         | 191 ++++++++++++++++
 src/test/regress/sql/matview.sql              | 108 +++++++++
 11 files changed, 582 insertions(+), 89 deletions(-)

diff --git a/doc/src/sgml/ref/create_materialized_view.sgml b/doc/src/sgml/ref/create_materialized_view.sgml
index 0d2fea2b97..b5a8e3441a 100644
--- a/doc/src/sgml/ref/create_materialized_view.sgml
+++ b/doc/src/sgml/ref/create_materialized_view.sgml
@@ -21,7 +21,7 @@ PostgreSQL documentation
 
  <refsynopsisdiv>
 <synopsis>
-CREATE MATERIALIZED VIEW [ IF NOT EXISTS ] <replaceable>table_name</replaceable>
+CREATE [ OR REPLACE ] MATERIALIZED VIEW [ IF NOT EXISTS ] <replaceable>table_name</replaceable>
     [ (<replaceable>column_name</replaceable> [, ...] ) ]
     [ USING <replaceable class="parameter">method</replaceable> ]
     [ WITH ( <replaceable class="parameter">storage_parameter</replaceable> [= <replaceable class="parameter">value</replaceable>] [, ... ] ) ]
@@ -60,6 +60,17 @@ CREATE MATERIALIZED VIEW [ IF NOT EXISTS ] <replaceable>table_name</replaceable>
   <title>Parameters</title>
 
   <variablelist>
+   <varlistentry>
+    <term><literal>OR REPLACE</literal></term>
+    <listitem>
+     <para>
+      Replaces a materialized view if it already exists.
+      Specifying <literal>OR REPLACE</literal> together with
+      <literal>IF NOT EXISTS</literal> is an error.
+     </para>
+    </listitem>
+   </varlistentry>
+
    <varlistentry>
     <term><literal>IF NOT EXISTS</literal></term>
     <listitem>
@@ -67,7 +78,7 @@ CREATE MATERIALIZED VIEW [ IF NOT EXISTS ] <replaceable>table_name</replaceable>
       Do not throw an error if a materialized view with the same name already
       exists. A notice is issued in this case.  Note that there is no guarantee
       that the existing materialized view is anything like the one that would
-      have been created.
+      have been created, unless you use <literal>OR REPLACE</literal> instead.
      </para>
     </listitem>
    </varlistentry>
diff --git a/src/backend/commands/createas.c b/src/backend/commands/createas.c
index 2c8a93b6e5..c5d78252a1 100644
--- a/src/backend/commands/createas.c
+++ b/src/backend/commands/createas.c
@@ -79,55 +79,151 @@ static void intorel_destroy(DestReceiver *self);
 static ObjectAddress
 create_ctas_internal(List *attrList, IntoClause *into)
 {
-	CreateStmt *create = makeNode(CreateStmt);
-	bool		is_matview;
+	bool		is_matview,
+				replace = false;
 	char		relkind;
-	Datum		toast_options;
-	static char *validnsps[] = HEAP_RELOPT_NAMESPACES;
+	Oid			matviewOid = InvalidOid;
 	ObjectAddress intoRelationAddr;
 
 	/* This code supports both CREATE TABLE AS and CREATE MATERIALIZED VIEW */
 	is_matview = (into->viewQuery != NULL);
 	relkind = is_matview ? RELKIND_MATVIEW : RELKIND_RELATION;
 
-	/*
-	 * Create the target relation by faking up a CREATE TABLE parsetree and
-	 * passing it to DefineRelation.
-	 */
-	create->relation = into->rel;
-	create->tableElts = attrList;
-	create->inhRelations = NIL;
-	create->ofTypename = NULL;
-	create->constraints = NIL;
-	create->options = into->options;
-	create->oncommit = into->onCommit;
-	create->tablespacename = into->tableSpaceName;
-	create->if_not_exists = false;
-	create->accessMethod = into->accessMethod;
+	/* Check if an existing materialized view needs to be replaced. */
+	if (is_matview)
+	{
+		LOCKMODE	lockmode;
 
-	/*
-	 * Create the relation.  (This will error out if there's an existing view,
-	 * so we don't need more code to complain if "replace" is false.)
-	 */
-	intoRelationAddr = DefineRelation(create, relkind, InvalidOid, NULL, NULL);
+		lockmode = into->replace ? AccessExclusiveLock : NoLock;
+		(void) RangeVarGetAndCheckCreationNamespace(into->rel, lockmode,
+													&matviewOid);
+		replace = OidIsValid(matviewOid) && into->replace;
+	}
 
-	/*
-	 * If necessary, create a TOAST table for the target table.  Note that
-	 * NewRelationCreateToastTable ends with CommandCounterIncrement(), so
-	 * that the TOAST table will be visible for insertion.
-	 */
-	CommandCounterIncrement();
+	if (is_matview && replace)
+	{
+		Relation	rel;
+		List	   *atcmds = NIL;
+		AlterTableCmd *atcmd;
+		TupleDesc	descriptor;
+
+		rel = relation_open(matviewOid, NoLock);
+
+		if (rel->rd_rel->relkind != RELKIND_MATVIEW)
+			ereport(ERROR,
+					errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					errmsg("\"%s\" is not a materialized view",
+						   RelationGetRelationName(rel)));
+
+		CheckTableNotInUse(rel, "CREATE OR REPLACE MATERIALIZED VIEW");
+
+		descriptor = BuildDescForRelation(attrList);
+		checkViewColumns(descriptor, rel->rd_att, true);
+
+		/* Add new attributes via ALTER TABLE. */
+		if (list_length(attrList) > rel->rd_att->natts)
+		{
+			ListCell   *c;
+			int			skip = rel->rd_att->natts;
+
+			foreach(c, attrList)
+			{
+				if (skip > 0)
+				{
+					skip--;
+					continue;
+				}
+				atcmd = makeNode(AlterTableCmd);
+				atcmd->subtype = AT_AddColumnToView;
+				atcmd->def = (Node *) lfirst(c);
+				atcmds = lappend(atcmds, atcmd);
+			}
+		}
+
+		/* Set access method via ALTER TABLE. */
+		if (into->accessMethod != NULL)
+		{
+			atcmd = makeNode(AlterTableCmd);
+			atcmd->subtype = AT_SetAccessMethod;
+			atcmd->name = into->accessMethod;
+			atcmds = lappend(atcmds, atcmd);
+		}
+
+		/* Set tablespace via ALTER TABLE. */
+		if (into->tableSpaceName != NULL)
+		{
+			atcmd = makeNode(AlterTableCmd);
+			atcmd->subtype = AT_SetTableSpace;
+			atcmd->name = into->tableSpaceName;
+			atcmds = lappend(atcmds, atcmd);
+		}
+
+		/* Set storage parameters via ALTER TABLE. */
+		if (into->options != NIL)
+		{
+			atcmd = makeNode(AlterTableCmd);
+			atcmd->subtype = AT_ReplaceRelOptions;
+			atcmd->def = (Node *) into->options;
+			atcmds = lappend(atcmds, atcmd);
+		}
+
+		if (atcmds != NIL)
+		{
+			AlterTableInternal(matviewOid, atcmds, true);
+			CommandCounterIncrement();
+		}
+
+		relation_close(rel, NoLock);
+
+		ObjectAddressSet(intoRelationAddr, RelationRelationId, matviewOid);
+	}
+	else
+	{
+		CreateStmt *create = makeNode(CreateStmt);
+		Datum		toast_options;
+		static char *validnsps[] = HEAP_RELOPT_NAMESPACES;
+
+		/*
+		 * Create the target relation by faking up a CREATE TABLE parsetree
+		 * and passing it to DefineRelation.
+		 */
+		create->relation = into->rel;
+		create->tableElts = attrList;
+		create->inhRelations = NIL;
+		create->ofTypename = NULL;
+		create->constraints = NIL;
+		create->options = into->options;
+		create->oncommit = into->onCommit;
+		create->tablespacename = into->tableSpaceName;
+		create->if_not_exists = false;
+		create->accessMethod = into->accessMethod;
+
+		/*
+		 * Create the relation.  (This will error out if there's an existing
+		 * view, so we don't need more code to complain if "replace" is
+		 * false.)
+		 */
+		intoRelationAddr = DefineRelation(create, relkind, InvalidOid, NULL,
+										  NULL);
 
-	/* parse and validate reloptions for the toast table */
-	toast_options = transformRelOptions((Datum) 0,
-										create->options,
-										"toast",
-										validnsps,
-										true, false);
+		/*
+		 * If necessary, create a TOAST table for the target table.  Note that
+		 * NewRelationCreateToastTable ends with CommandCounterIncrement(), so
+		 * that the TOAST table will be visible for insertion.
+		 */
+		CommandCounterIncrement();
+
+		/* parse and validate reloptions for the toast table */
+		toast_options = transformRelOptions((Datum) 0,
+											create->options,
+											"toast",
+											validnsps,
+											true, false);
 
-	(void) heap_reloptions(RELKIND_TOASTVALUE, toast_options, true);
+		(void) heap_reloptions(RELKIND_TOASTVALUE, toast_options, true);
 
-	NewRelationCreateToastTable(intoRelationAddr.objectId, toast_options);
+		NewRelationCreateToastTable(intoRelationAddr.objectId, toast_options);
+	}
 
 	/* Create the "view" part of a materialized view. */
 	if (is_matview)
@@ -135,7 +231,7 @@ create_ctas_internal(List *attrList, IntoClause *into)
 		/* StoreViewQuery scribbles on tree, so make a copy */
 		Query	   *query = (Query *) copyObject(into->viewQuery);
 
-		StoreViewQuery(intoRelationAddr.objectId, query, false);
+		StoreViewQuery(intoRelationAddr.objectId, query, replace);
 		CommandCounterIncrement();
 	}
 
@@ -234,7 +330,26 @@ ExecCreateTableAs(ParseState *pstate, CreateTableAsStmt *stmt,
 
 	/* Check if the relation exists or not */
 	if (CreateTableAsRelExists(stmt))
+	{
+		/* An existing materialized view can be replaced. */
+		if (is_matview && into->replace)
+		{
+			RefreshMatViewStmt *refresh;
+
+			/* Change the relation to match the new query and other options. */
+			(void) create_ctas_nodata(query->targetList, into);
+
+			/* Refresh the materialized view with a fake statement. */
+			refresh = makeNode(RefreshMatViewStmt);
+			refresh->relation = into->rel;
+			refresh->skipData = into->skipData;
+			refresh->concurrent = false;
+
+			return ExecRefreshMatView(refresh, NULL, NULL, NULL);
+		}
+
 		return InvalidObjectAddress;
+	}
 
 	/*
 	 * Create the tuple receiver object and insert info it will need
@@ -395,14 +510,15 @@ CreateTableAsRelExists(CreateTableAsStmt *ctas)
 	oldrelid = get_relname_relid(into->rel->relname, nspid);
 	if (OidIsValid(oldrelid))
 	{
-		if (!ctas->if_not_exists)
+		if (!ctas->if_not_exists && !into->replace)
 			ereport(ERROR,
 					(errcode(ERRCODE_DUPLICATE_TABLE),
 					 errmsg("relation \"%s\" already exists",
 							into->rel->relname)));
 
 		/*
-		 * The relation exists and IF NOT EXISTS has been specified.
+		 * The relation exists and IF NOT EXISTS or OR REPLACE has been
+		 * specified.
 		 *
 		 * If we are in an extension script, insist that the pre-existing
 		 * object be a member of the extension, to avoid security risks.
@@ -410,11 +526,12 @@ CreateTableAsRelExists(CreateTableAsStmt *ctas)
 		ObjectAddressSet(address, RelationRelationId, oldrelid);
 		checkMembershipInCurrentExtension(&address);
 
-		/* OK to skip */
-		ereport(NOTICE,
-				(errcode(ERRCODE_DUPLICATE_TABLE),
-				 errmsg("relation \"%s\" already exists, skipping",
-						into->rel->relname)));
+		if (ctas->if_not_exists)
+			/* OK to skip */
+			ereport(NOTICE,
+					(errcode(ERRCODE_DUPLICATE_TABLE),
+					 errmsg("relation \"%s\" already exists, skipping",
+							into->rel->relname)));
 		return true;
 	}
 
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 0b2a52463f..5274daaf3a 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -4484,7 +4484,7 @@ AlterTableGetLockLevel(List *cmds)
 				 * Subcommands that may be visible to concurrent SELECTs
 				 */
 			case AT_DropColumn: /* change visible to SELECT */
-			case AT_AddColumnToView:	/* CREATE VIEW */
+			case AT_AddColumnToView:	/* via CREATE OR REPLACE [MATERIALIZED] VIEW */
 			case AT_DropOids:	/* used to equiv to DropColumn */
 			case AT_EnableAlwaysRule:	/* may change SELECT rules */
 			case AT_EnableReplicaRule:	/* may change SELECT rules */
@@ -4795,8 +4795,8 @@ ATPrepCmd(List **wqueue, Relation rel, AlterTableCmd *cmd,
 			/* Recursion occurs during execution phase */
 			pass = AT_PASS_ADD_COL;
 			break;
-		case AT_AddColumnToView:	/* add column via CREATE OR REPLACE VIEW */
-			ATSimplePermissions(cmd->subtype, rel, ATT_VIEW);
+		case AT_AddColumnToView:	/* via CREATE OR REPLACE [MATERIALIZED] VIEW */
+			ATSimplePermissions(cmd->subtype, rel, ATT_VIEW | ATT_MATVIEW);
 			ATPrepAddColumn(wqueue, rel, recurse, recursing, true, cmd,
 							lockmode, context);
 			/* Recursion occurs during execution phase */
@@ -5220,7 +5220,7 @@ ATExecCmd(List **wqueue, AlteredTableInfo *tab,
 	switch (cmd->subtype)
 	{
 		case AT_AddColumn:		/* ADD COLUMN */
-		case AT_AddColumnToView:	/* add column via CREATE OR REPLACE VIEW */
+		case AT_AddColumnToView:	/* via CREATE OR REPLACE [MATERIALIZED] VIEW */
 			address = ATExecAddColumn(wqueue, tab, rel, &cmd,
 									  cmd->recurse, false,
 									  lockmode, cur_pass, context);
diff --git a/src/backend/commands/view.c b/src/backend/commands/view.c
index fdad833832..76532aa35d 100644
--- a/src/backend/commands/view.c
+++ b/src/backend/commands/view.c
@@ -30,8 +30,6 @@
 #include "utils/lsyscache.h"
 #include "utils/rel.h"
 
-static void checkViewColumns(TupleDesc newdesc, TupleDesc olddesc);
-
 /*---------------------------------------------------------------------
  * DefineVirtualRelation
  *
@@ -130,7 +128,7 @@ DefineVirtualRelation(RangeVar *relation, List *tlist, bool replace,
 		 * column list.
 		 */
 		descriptor = BuildDescForRelation(attrList);
-		checkViewColumns(descriptor, rel->rd_att);
+		checkViewColumns(descriptor, rel->rd_att, false);
 
 		/*
 		 * If new attributes have been added, we must add pg_attribute entries
@@ -263,15 +261,22 @@ DefineVirtualRelation(RangeVar *relation, List *tlist, bool replace,
  * added to generate specific complaints.  Also, we allow the new view to have
  * more columns than the old.
  */
-static void
-checkViewColumns(TupleDesc newdesc, TupleDesc olddesc)
+void
+checkViewColumns(TupleDesc newdesc, TupleDesc olddesc, bool is_matview)
 {
 	int			i;
 
 	if (newdesc->natts < olddesc->natts)
-		ereport(ERROR,
-				(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
-				 errmsg("cannot drop columns from view")));
+	{
+		if (is_matview)
+			ereport(ERROR,
+					errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+					errmsg("cannot drop columns from materialized view"));
+		else
+			ereport(ERROR,
+					(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+					 errmsg("cannot drop columns from view")));
+	}
 
 	for (i = 0; i < olddesc->natts; i++)
 	{
@@ -280,17 +285,34 @@ checkViewColumns(TupleDesc newdesc, TupleDesc olddesc)
 
 		/* XXX msg not right, but we don't support DROP COL on view anyway */
 		if (newattr->attisdropped != oldattr->attisdropped)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
-					 errmsg("cannot drop columns from view")));
+		{
+			if (is_matview)
+				ereport(ERROR,
+						errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						errmsg("cannot drop columns from materialized view"));
+			else
+				ereport(ERROR,
+						(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						 errmsg("cannot drop columns from view")));
+		}
 
 		if (strcmp(NameStr(newattr->attname), NameStr(oldattr->attname)) != 0)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
-					 errmsg("cannot change name of view column \"%s\" to \"%s\"",
-							NameStr(oldattr->attname),
-							NameStr(newattr->attname)),
-					 errhint("Use ALTER VIEW ... RENAME COLUMN ... to change name of view column instead.")));
+		{
+			if (is_matview)
+				ereport(ERROR,
+						errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						errmsg("cannot change name of materialized view column \"%s\" to \"%s\"",
+							   NameStr(oldattr->attname),
+							   NameStr(newattr->attname)),
+						errhint("Use ALTER MATERIALIZED VIEW ... RENAME COLUMN ... to change name of materialized view column instead."));
+			else
+				ereport(ERROR,
+						(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						 errmsg("cannot change name of view column \"%s\" to \"%s\"",
+								NameStr(oldattr->attname),
+								NameStr(newattr->attname)),
+						 errhint("Use ALTER VIEW ... RENAME COLUMN ... to change name of view column instead.")));
+		}
 
 		/*
 		 * We cannot allow type, typmod, or collation to change, since these
@@ -299,26 +321,48 @@ checkViewColumns(TupleDesc newdesc, TupleDesc olddesc)
 		 */
 		if (newattr->atttypid != oldattr->atttypid ||
 			newattr->atttypmod != oldattr->atttypmod)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
-					 errmsg("cannot change data type of view column \"%s\" from %s to %s",
-							NameStr(oldattr->attname),
-							format_type_with_typemod(oldattr->atttypid,
-													 oldattr->atttypmod),
-							format_type_with_typemod(newattr->atttypid,
-													 newattr->atttypmod))));
+		{
+			if (is_matview)
+				ereport(ERROR,
+						errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						errmsg("cannot change data type of materialized view column \"%s\" from %s to %s",
+							   NameStr(oldattr->attname),
+							   format_type_with_typemod(oldattr->atttypid,
+														oldattr->atttypmod),
+							   format_type_with_typemod(newattr->atttypid,
+														newattr->atttypmod)));
+			else
+				ereport(ERROR,
+						(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						 errmsg("cannot change data type of view column \"%s\" from %s to %s",
+								NameStr(oldattr->attname),
+								format_type_with_typemod(oldattr->atttypid,
+														 oldattr->atttypmod),
+								format_type_with_typemod(newattr->atttypid,
+														 newattr->atttypmod))));
+		}
 
 		/*
 		 * At this point, attcollations should be both valid or both invalid,
 		 * so applying get_collation_name unconditionally should be fine.
 		 */
 		if (newattr->attcollation != oldattr->attcollation)
-			ereport(ERROR,
-					(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
-					 errmsg("cannot change collation of view column \"%s\" from \"%s\" to \"%s\"",
-							NameStr(oldattr->attname),
-							get_collation_name(oldattr->attcollation),
-							get_collation_name(newattr->attcollation))));
+		{
+			if (is_matview)
+				ereport(ERROR,
+						errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						errmsg("cannot change collation of materialized view column \"%s\" from \"%s\" to \"%s\"",
+							   NameStr(oldattr->attname),
+							   get_collation_name(oldattr->attcollation),
+							   get_collation_name(newattr->attcollation)));
+			else
+				ereport(ERROR,
+						(errcode(ERRCODE_INVALID_TABLE_DEFINITION),
+						 errmsg("cannot change collation of view column \"%s\" from \"%s\" to \"%s\"",
+								NameStr(oldattr->attname),
+								get_collation_name(oldattr->attcollation),
+								get_collation_name(newattr->attcollation))));
+		}
 	}
 
 	/*
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index a043fd4c66..ed477806d1 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -4831,6 +4831,21 @@ CreateMatViewStmt:
 					$8->skipData = !($11);
 					$$ = (Node *) ctas;
 				}
+		| CREATE OR REPLACE OptNoLog MATERIALIZED VIEW create_mv_target AS SelectStmt opt_with_data
+				{
+					CreateTableAsStmt *ctas = makeNode(CreateTableAsStmt);
+
+					ctas->query = $9;
+					ctas->into = $7;
+					ctas->objtype = OBJECT_MATVIEW;
+					ctas->is_select_into = false;
+					ctas->if_not_exists = false;
+					/* cram additional flags into the IntoClause */
+					$7->rel->relpersistence = $4;
+					$7->skipData = !($10);
+					$7->replace = true;
+					$$ = (Node *) ctas;
+				}
 		;
 
 create_mv_target:
diff --git a/src/bin/psql/tab-complete.c b/src/bin/psql/tab-complete.c
index 891face1b6..f3291e79d8 100644
--- a/src/bin/psql/tab-complete.c
+++ b/src/bin/psql/tab-complete.c
@@ -1811,7 +1811,7 @@ psql_completion(const char *text, int start, int end)
 	/* complete with something you can create or replace */
 	else if (TailMatches("CREATE", "OR", "REPLACE"))
 		COMPLETE_WITH("FUNCTION", "PROCEDURE", "LANGUAGE", "RULE", "VIEW",
-					  "AGGREGATE", "TRANSFORM", "TRIGGER");
+					  "AGGREGATE", "TRANSFORM", "TRIGGER", "MATERIALIZED VIEW");
 
 /* DROP, but not DROP embedded in other commands */
 	/* complete with something you can drop */
@@ -3619,13 +3619,16 @@ psql_completion(const char *text, int start, int end)
 		COMPLETE_WITH("SELECT");
 
 /* CREATE MATERIALIZED VIEW */
-	else if (Matches("CREATE", "MATERIALIZED"))
+	else if (Matches("CREATE", "MATERIALIZED") ||
+			 Matches("CREATE", "OR", "REPLACE", "MATERIALIZED"))
 		COMPLETE_WITH("VIEW");
-	/* Complete CREATE MATERIALIZED VIEW <name> with AS */
-	else if (Matches("CREATE", "MATERIALIZED", "VIEW", MatchAny))
+	/* Complete CREATE [ OR REPLACE ] MATERIALIZED VIEW <name> with AS */
+	else if (Matches("CREATE", "MATERIALIZED", "VIEW", MatchAny) ||
+			 Matches("CREATE", "OR", "REPLACE", "MATERIALIZED", "VIEW", MatchAny))
 		COMPLETE_WITH("AS");
-	/* Complete "CREATE MATERIALIZED VIEW <sth> AS with "SELECT" */
-	else if (Matches("CREATE", "MATERIALIZED", "VIEW", MatchAny, "AS"))
+	/* Complete "CREATE [ OR REPLACE ] MATERIALIZED VIEW <sth> AS with "SELECT" */
+	else if (Matches("CREATE", "MATERIALIZED", "VIEW", MatchAny, "AS") ||
+			 Matches("CREATE", "OR", "REPLACE", "MATERIALIZED", "VIEW", MatchAny, "AS"))
 		COMPLETE_WITH("SELECT");
 
 /* CREATE EVENT TRIGGER */
diff --git a/src/include/commands/view.h b/src/include/commands/view.h
index d2d8588989..7eacdaaceb 100644
--- a/src/include/commands/view.h
+++ b/src/include/commands/view.h
@@ -22,4 +22,7 @@ extern ObjectAddress DefineView(ViewStmt *stmt, const char *queryString,
 
 extern void StoreViewQuery(Oid viewOid, Query *viewParse, bool replace);
 
+extern void checkViewColumns(TupleDesc newdesc, TupleDesc olddesc,
+							 bool is_matview);
+
 #endif							/* VIEW_H */
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
index 85a62b538e..cbc4b608e3 100644
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -2351,7 +2351,7 @@ typedef struct AlterTableStmt
 typedef enum AlterTableType
 {
 	AT_AddColumn,				/* add column */
-	AT_AddColumnToView,			/* implicitly via CREATE OR REPLACE VIEW */
+	AT_AddColumnToView,			/* implicitly via CREATE OR REPLACE [MATERIALIZED] VIEW */
 	AT_ColumnDefault,			/* alter column default */
 	AT_CookedColumnDefault,		/* add a pre-cooked column default */
 	AT_DropNotNull,				/* alter column drop not null */
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
index ea47652adb..4b0ee5d10d 100644
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -168,6 +168,7 @@ typedef struct IntoClause
 	/* materialized view's SELECT query */
 	Node	   *viewQuery pg_node_attr(query_jumble_ignore);
 	bool		skipData;		/* true for WITH NO DATA */
+	bool		replace;		/* replace existing matview? */
 } IntoClause;
 
 
diff --git a/src/test/regress/expected/matview.out b/src/test/regress/expected/matview.out
index 038ab73517..e2e2a13396 100644
--- a/src/test/regress/expected/matview.out
+++ b/src/test/regress/expected/matview.out
@@ -694,3 +694,194 @@ NOTICE:  relation "matview_ine_tab" already exists, skipping
 (0 rows)
 
 DROP MATERIALIZED VIEW matview_ine_tab;
+--
+-- test CREATE OR REPLACE MATERIALIZED VIEW
+--
+-- matview does not already exist
+DROP MATERIALIZED VIEW IF EXISTS mvtest_replace;
+NOTICE:  materialized view "mvtest_replace" does not exist, skipping
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 1 AS a;
+SELECT * FROM mvtest_replace;
+ a 
+---
+ 1
+(1 row)
+
+-- replace query with data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 2 AS a;
+SELECT * FROM mvtest_replace;
+ a 
+---
+ 2
+(1 row)
+
+-- replace query without data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 3 AS a
+  WITH NO DATA;
+SELECT * FROM mvtest_replace; -- error: not populated
+ERROR:  materialized view "mvtest_replace" has not been populated
+HINT:  Use the REFRESH MATERIALIZED VIEW command.
+REFRESH MATERIALIZED VIEW mvtest_replace;
+SELECT * FROM mvtest_replace;
+ a 
+---
+ 3
+(1 row)
+
+-- add column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 4 AS a, 1 b;
+SELECT * FROM mvtest_replace;
+ a | b 
+---+---
+ 4 | 1
+(1 row)
+
+-- replace table options
+SELECT m.*, c.relname, c.reloptions, s.spcname, a.amname
+  FROM mvtest_replace m
+    CROSS JOIN pg_class c
+    LEFT JOIN pg_tablespace s ON s.oid = c.reltablespace
+    LEFT JOIN pg_am a ON a.oid = c.relam
+  WHERE c.relname = 'mvtest_replace';
+ a | b |    relname     | reloptions | spcname | amname 
+---+---+----------------+------------+---------+--------
+ 4 | 1 | mvtest_replace |            |         | heap
+(1 row)
+
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace
+  USING heap2
+  WITH (fillfactor = 50)
+  TABLESPACE regress_tblspace
+  AS SELECT 5 AS a, 1 AS b;
+SELECT m.*, c.relname, c.reloptions, s.spcname, a.amname
+  FROM mvtest_replace m
+    CROSS JOIN pg_class c
+    LEFT JOIN pg_tablespace s ON s.oid = c.reltablespace
+    LEFT JOIN pg_am a ON a.oid = c.relam
+  WHERE c.relname = 'mvtest_replace';
+ a | b |    relname     |   reloptions    |     spcname      | amname 
+---+---+----------------+-----------------+------------------+--------
+ 5 | 1 | mvtest_replace | {fillfactor=50} | regress_tblspace | heap2
+(1 row)
+
+-- can replace matview that has a dependent view
+CREATE VIEW mvtest_replace_v AS
+  SELECT * FROM mvtest_replace;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 6 AS a, 1 AS b;
+SELECT * FROM mvtest_replace, mvtest_replace_v;
+ a | b | a | b 
+---+---+---+---
+ 6 | 1 | 6 | 1
+(1 row)
+
+DROP VIEW mvtest_replace_v;
+-- index gets rebuilt when replacing with data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 7 AS a, 1 AS b;
+CREATE UNIQUE INDEX ON mvtest_replace (b);
+SELECT * FROM mvtest_replace;
+ a | b 
+---+---
+ 7 | 1
+(1 row)
+
+SET enable_seqscan = off; -- force index scan
+EXPLAIN (COSTS OFF) SELECT * FROM mvtest_replace WHERE b = 1;
+                       QUERY PLAN                        
+---------------------------------------------------------
+ Index Scan using mvtest_replace_b_idx on mvtest_replace
+   Index Cond: (b = 1)
+(2 rows)
+
+SELECT * FROM mvtest_replace WHERE b = 1;
+ a | b 
+---+---
+ 7 | 1
+(1 row)
+
+RESET enable_seqscan;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 8 AS a, 1 AS b;
+SET enable_seqscan = off; -- force index scan
+EXPLAIN (COSTS OFF) SELECT * FROM mvtest_replace WHERE b = 1;
+                       QUERY PLAN                        
+---------------------------------------------------------
+ Index Scan using mvtest_replace_b_idx on mvtest_replace
+   Index Cond: (b = 1)
+(2 rows)
+
+SELECT * FROM mvtest_replace WHERE b = 1;
+ a | b 
+---+---
+ 8 | 1
+(1 row)
+
+RESET enable_seqscan;
+-- cannot change column data type
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 9 AS a, 'x' AS b; -- error
+ERROR:  cannot change data type of materialized view column "b" from integer to text
+SELECT * FROM mvtest_replace;
+ a | b 
+---+---
+ 8 | 1
+(1 row)
+
+-- cannot rename column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 10 AS a, 1 AS b2; -- error
+ERROR:  cannot change name of materialized view column "b" to "b2"
+HINT:  Use ALTER MATERIALIZED VIEW ... RENAME COLUMN ... to change name of materialized view column instead.
+SELECT * FROM mvtest_replace;
+ a | b 
+---+---
+ 8 | 1
+(1 row)
+
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 11 AS a, 1 AS b, 'y' COLLATE "C" AS c;
+SELECT * FROM mvtest_replace;
+ a  | b | c 
+----+---+---
+ 11 | 1 | y
+(1 row)
+
+-- cannot change column collation
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 12 AS a, 1 AS b, 'x' COLLATE "POSIX" AS c; -- error
+ERROR:  cannot change collation of materialized view column "c" from "C" to "POSIX"
+SELECT * FROM mvtest_replace;
+ a  | b | c 
+----+---+---
+ 11 | 1 | y
+(1 row)
+
+-- cannot drop column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 13 AS a, 1 AS b; -- error
+ERROR:  cannot drop columns from materialized view
+SELECT * FROM mvtest_replace;
+ a  | b | c 
+----+---+---
+ 11 | 1 | y
+(1 row)
+
+-- must target a matview
+CREATE VIEW mvtest_not_mv AS
+  SELECT 1 AS a;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_not_mv AS
+  SELECT 1 AS a; -- error
+ERROR:  "mvtest_not_mv" is not a materialized view
+DROP VIEW mvtest_not_mv;
+-- cannot use OR REPLACE with IF NOT EXISTS
+CREATE OR REPLACE MATERIALIZED VIEW IF NOT EXISTS mvtest_replace AS
+  SELECT 1 AS a;
+ERROR:  syntax error at or near "NOT"
+LINE 1: CREATE OR REPLACE MATERIALIZED VIEW IF NOT EXISTS mvtest_rep...
+                                               ^
+DROP MATERIALIZED VIEW mvtest_replace;
diff --git a/src/test/regress/sql/matview.sql b/src/test/regress/sql/matview.sql
index b74ee305e0..c12f0243c9 100644
--- a/src/test/regress/sql/matview.sql
+++ b/src/test/regress/sql/matview.sql
@@ -314,3 +314,111 @@ EXPLAIN (ANALYZE, COSTS OFF, SUMMARY OFF, TIMING OFF)
   CREATE MATERIALIZED VIEW IF NOT EXISTS matview_ine_tab AS
     SELECT 1 / 0 WITH NO DATA; -- ok
 DROP MATERIALIZED VIEW matview_ine_tab;
+
+--
+-- test CREATE OR REPLACE MATERIALIZED VIEW
+--
+
+-- matview does not already exist
+DROP MATERIALIZED VIEW IF EXISTS mvtest_replace;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 1 AS a;
+SELECT * FROM mvtest_replace;
+
+-- replace query with data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 2 AS a;
+SELECT * FROM mvtest_replace;
+
+-- replace query without data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 3 AS a
+  WITH NO DATA;
+SELECT * FROM mvtest_replace; -- error: not populated
+REFRESH MATERIALIZED VIEW mvtest_replace;
+SELECT * FROM mvtest_replace;
+
+-- add column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 4 AS a, 1 b;
+SELECT * FROM mvtest_replace;
+
+-- replace table options
+SELECT m.*, c.relname, c.reloptions, s.spcname, a.amname
+  FROM mvtest_replace m
+    CROSS JOIN pg_class c
+    LEFT JOIN pg_tablespace s ON s.oid = c.reltablespace
+    LEFT JOIN pg_am a ON a.oid = c.relam
+  WHERE c.relname = 'mvtest_replace';
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace
+  USING heap2
+  WITH (fillfactor = 50)
+  TABLESPACE regress_tblspace
+  AS SELECT 5 AS a, 1 AS b;
+SELECT m.*, c.relname, c.reloptions, s.spcname, a.amname
+  FROM mvtest_replace m
+    CROSS JOIN pg_class c
+    LEFT JOIN pg_tablespace s ON s.oid = c.reltablespace
+    LEFT JOIN pg_am a ON a.oid = c.relam
+  WHERE c.relname = 'mvtest_replace';
+
+-- can replace matview that has a dependent view
+CREATE VIEW mvtest_replace_v AS
+  SELECT * FROM mvtest_replace;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 6 AS a, 1 AS b;
+SELECT * FROM mvtest_replace, mvtest_replace_v;
+DROP VIEW mvtest_replace_v;
+
+-- index gets rebuilt when replacing with data
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 7 AS a, 1 AS b;
+CREATE UNIQUE INDEX ON mvtest_replace (b);
+SELECT * FROM mvtest_replace;
+SET enable_seqscan = off; -- force index scan
+EXPLAIN (COSTS OFF) SELECT * FROM mvtest_replace WHERE b = 1;
+SELECT * FROM mvtest_replace WHERE b = 1;
+RESET enable_seqscan;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 8 AS a, 1 AS b;
+SET enable_seqscan = off; -- force index scan
+EXPLAIN (COSTS OFF) SELECT * FROM mvtest_replace WHERE b = 1;
+SELECT * FROM mvtest_replace WHERE b = 1;
+RESET enable_seqscan;
+
+-- cannot change column data type
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 9 AS a, 'x' AS b; -- error
+SELECT * FROM mvtest_replace;
+
+-- cannot rename column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 10 AS a, 1 AS b2; -- error
+SELECT * FROM mvtest_replace;
+
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 11 AS a, 1 AS b, 'y' COLLATE "C" AS c;
+SELECT * FROM mvtest_replace;
+
+-- cannot change column collation
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 12 AS a, 1 AS b, 'x' COLLATE "POSIX" AS c; -- error
+SELECT * FROM mvtest_replace;
+
+-- cannot drop column
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_replace AS
+  SELECT 13 AS a, 1 AS b; -- error
+SELECT * FROM mvtest_replace;
+
+-- must target a matview
+CREATE VIEW mvtest_not_mv AS
+  SELECT 1 AS a;
+CREATE OR REPLACE MATERIALIZED VIEW mvtest_not_mv AS
+  SELECT 1 AS a; -- error
+DROP VIEW mvtest_not_mv;
+
+-- cannot use OR REPLACE with IF NOT EXISTS
+CREATE OR REPLACE MATERIALIZED VIEW IF NOT EXISTS mvtest_replace AS
+  SELECT 1 AS a;
+
+DROP MATERIALIZED VIEW mvtest_replace;
-- 
2.45.2


--nrp556lrhtfteavd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0002-Deprecate-CREATE-MATERIALIZED-VIEW-IF-NOT-EXISTS.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread

* [PATCH v47 4/9] repack code cleanups
@ 2026-03-31 14:48 Álvaro Herrera <[email protected]>
  0 siblings, 0 replies; 197+ messages in thread

From: Álvaro Herrera @ 2026-03-31 14:48 UTC (permalink / raw)

---
 src/backend/commands/cluster.c | 163 +++++++++++----------------------
 1 file changed, 51 insertions(+), 112 deletions(-)

diff --git a/src/backend/commands/cluster.c b/src/backend/commands/cluster.c
index fd0bfacaae6..c0220cb3380 100644
--- a/src/backend/commands/cluster.c
+++ b/src/backend/commands/cluster.c
@@ -2459,18 +2459,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	TupleTableSlot *spilled_tuple;
 	TupleTableSlot *old_update_tuple;
 	TupleTableSlot *ondisk_tuple;
-	MemoryContext apply_cxt;
 	bool		have_old_tuple = false;
-
-	/*
-	 * Use a separate memory context for the tuples and any memory needed to
-	 * process them.
-	 *
-	 * XXX would this be better with GenerationContextCreate?
-	 */
-	apply_cxt = AllocSetContextCreate(TopTransactionContext,
-									  "REPACK - apply",
-									  ALLOCSET_DEFAULT_SIZES);
+	MemoryContext	oldcxt;
 
 	spilled_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 											 &TTSOpsVirtual);
@@ -2479,6 +2469,8 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 	old_update_tuple = MakeSingleTupleTableSlot(RelationGetDescr(rel),
 												&TTSOpsVirtual);
 
+	oldcxt = MemoryContextSwitchTo(GetPerTupleMemoryContext(chgcxt->cc_estate));
+
 	while (true)
 	{
 		size_t		nread;
@@ -2570,14 +2562,15 @@ apply_concurrent_changes(BufFile *file, ChangeContext *chgcxt)
 		else
 			elog(ERROR, "unrecognized kind of change: %d", kind);
 
-		MemoryContextReset(apply_cxt);
+		ResetPerTupleExprContext(chgcxt->cc_estate);
 	}
 
 	/* Cleanup. */
 	ExecDropSingleTupleTableSlot(spilled_tuple);
 	ExecDropSingleTupleTableSlot(ondisk_tuple);
 	ExecDropSingleTupleTableSlot(old_update_tuple);
-	MemoryContextDelete(apply_cxt);
+
+	MemoryContextSwitchTo(oldcxt);
 }
 
 /*
@@ -2588,27 +2581,17 @@ static void
 apply_concurrent_insert(Relation rel, TupleTableSlot *slot,
 						ChangeContext *chgcxt)
 {
-	List	   *recheck;
-
 	/* Put the tuple in the table, but make sure it won't be decoded */
 	table_tuple_insert(rel, slot, GetCurrentCommandId(true),
 					   HEAP_INSERT_NO_LOGICAL, NULL);
 
-	/*
-	 * Update indexes with this new tuple.  Because we're merely replaying an
-	 * action that already happened, we have no use for the recheck list of
-	 * indexes returned, so just free it.  XXX or maybe just leave it?
-	 */
-	recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-									chgcxt->cc_estate,
-									0,
-									slot,
-									NIL, NULL);
-	list_free(recheck);
-
+	/* Update indexes with this new tuple. */
+	ExecInsertIndexTuples(chgcxt->cc_rri,
+						  chgcxt->cc_estate,
+						  0,
+						  slot,
+						  NIL, NULL);
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_INSERTED, 1);
-
-	ResetPerTupleExprContext(chgcxt->cc_estate);
 }
 
 /*
@@ -2624,10 +2607,9 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 	TM_FailureData tmfd;
 	TU_UpdateIndexes update_indexes;
 	TM_Result	res;
-	List	   *recheck;
 
 	/*
-	 * Carry out the update, avoiding logical decoding info.
+	 * Carry out the update, skipping logical decoding for it.
 	 */
 	res = table_tuple_update(rel, &(ondisk_tuple->tts_tid), spilled_tuple,
 							 GetCurrentCommandId(true),
@@ -2646,13 +2628,11 @@ apply_concurrent_update(Relation rel, TupleTableSlot *spilled_tuple,
 
 		if (update_indexes == TU_Summarizing)
 			flags |= EIIT_ONLY_SUMMARIZING;
-		recheck = ExecInsertIndexTuples(chgcxt->cc_rri,
-										chgcxt->cc_estate,
-										flags,
-										spilled_tuple,
-										NIL, NULL);
-		list_free(recheck);
-		ResetPerTupleExprContext(chgcxt->cc_estate);
+		ExecInsertIndexTuples(chgcxt->cc_rri,
+							  chgcxt->cc_estate,
+							  flags,
+							  spilled_tuple,
+							  NIL, NULL);
 	}
 
 	pgstat_progress_incr_param(PROGRESS_REPACK_HEAP_TUPLES_UPDATED, 1);
@@ -2665,10 +2645,7 @@ apply_concurrent_delete(Relation rel, TupleTableSlot *slot)
 	TM_FailureData tmfd;
 
 	/*
-	 * Delete tuple from the new heap.
-	 *
-	 * Do it like in simple_heap_delete(), except for 'wal_logical' (and
-	 * except for 'wait').
+	 * Delete tuple from the new heap, skipping logical decoding for it.
 	 */
 	res = table_tuple_delete(rel, &(slot->tts_tid),
 							 GetCurrentCommandId(true),
@@ -3009,12 +2986,8 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	char		relpersistence;
 	bool		is_system_catalog;
 	Oid			ident_idx_new;
-	XLogRecPtr	wal_insert_ptr,
-				end_of_wal;
-	char		dummy_rec_data = '\0';
-	Relation   *ind_refs,
-			   *ind_refs_p;
-	int			nind;
+	XLogRecPtr	end_of_wal;
+	List	   *indexrels;
 	ChangeContext chgcxt;
 
 	/* Like in cluster_rel(). */
@@ -3041,24 +3014,23 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 */
 	ind_oids_new = build_new_indexes(NewHeap, OldHeap, ind_oids_old);
 
-	/* Find "identity index" on the new relation. */
+	/*
+	 * The identity index in the new relation appears in the same relative
+	 * position as the corresponding index in the old relation.  Find it.
+	 */
 	ident_idx_new = InvalidOid;
-	forboth(lc, ind_oids_old, lc2, ind_oids_new)
+	foreach_oid(ind_old, ind_oids_old)
 	{
-		Oid			ind_old = lfirst_oid(lc);
-		Oid			ind_new = lfirst_oid(lc2);
-
 		if (identIdx == ind_old)
 		{
-			ident_idx_new = ind_new;
+			ident_idx_new = list_nth_oid(ind_oids_new,
+										 foreach_current_index(ind_old));
 			break;
 		}
 	}
-
-	/* Should not happen, given our lock on the old relation. */
 	if (!OidIsValid(ident_idx_new))
-		ereport(ERROR,
-				errmsg("identity index missing on the new relation"));
+		elog(ERROR, "could not find index matching \"%s\" at the new relation",
+			 get_rel_name(identIdx));
 
 	/* Gather information to apply concurrent changes. */
 	initialize_change_context(&chgcxt, NewHeap, ident_idx_new);
@@ -3074,8 +3046,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * incomplete page; see GetInsertRecPtr), to minimize the amount of data
 	 * we need to flush while holding exclusive lock on the source table.
 	 */
-	wal_insert_ptr = GetInsertRecPtr();
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3094,10 +3065,9 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	/*
 	 * Lock all indexes now, not only the clustering one: all indexes need to
 	 * have their files swapped. While doing that, store their relation
-	 * references in an array, to handle predicate locks below.
+	 * references in a zero-terminated array, to handle predicate locks below.
 	 */
-	ind_refs_p = ind_refs = palloc_array(Relation, list_length(ind_oids_old));
-	nind = 0;
+	indexrels = NIL;
 	foreach_oid(ind_oid, ind_oids_old)
 	{
 		Relation	index;
@@ -3105,16 +3075,11 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 		index = index_open(ind_oid, AccessExclusiveLock);
 
 		/*
-		 * TODO 1) Do we need to check if ALTER INDEX was executed since the
-		 * new index was created in build_new_indexes()? 2) Specifically for
-		 * the clustering index, should check_index_is_clusterable() be called
-		 * here? (Not sure about the latter: ShareUpdateExclusiveLock on the
-		 * table probably blocks all commands that affect the result of
-		 * check_index_is_clusterable().)
+		 * Some things about the index may have changed before we locked the
+		 * index, such as ALTER INDEX RENAME.  We don't need to do anything
+		 * here to absorb those changes in the new index.
 		 */
-		*ind_refs_p = index;
-		ind_refs_p++;
-		nind++;
+		indexrels = lappend(indexrels, index);
 	}
 
 	/*
@@ -3128,40 +3093,18 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	 * Tuples and pages of the old heap will be gone, but the heap will stay.
 	 */
 	TransferPredicateLocksToHeapRelation(OldHeap);
-	/* The same for indexes. */
-	for (int i = 0; i < nind; i++)
+	foreach_ptr(RelationData, index, indexrels)
 	{
-		Relation	index = ind_refs[i];
-
 		TransferPredicateLocksToHeapRelation(index);
-
-		/*
-		 * References to indexes on the old relation are not needed anymore,
-		 * however locks stay till the end of the transaction.
-		 */
 		index_close(index, NoLock);
 	}
-	pfree(ind_refs);
+	list_free(indexrels);
 
 	/*
-	 * Flush anything we see in WAL, to make sure that all changes committed
-	 * while we were waiting for the exclusive lock are available for
-	 * decoding. This should not be necessary if all backends had
-	 * synchronous_commit set, but we can't rely on this setting.
-	 *
-	 * Unfortunately, GetInsertRecPtr() may lag behind the actual insert
-	 * position, and GetLastImportantRecPtr() points at the start of the last
-	 * record rather than at the end. Thus the simplest way to determine the
-	 * insert position is to insert a dummy record and use its LSN.
-	 *
-	 * XXX Consider using GetLastImportantRecPtr() and adding the size of the
-	 * last record (plus the total size of all the page headers the record
-	 * spans)?
+	 * Flush WAL again, to make sure that all changes committed while we were
+	 * waiting for the exclusive lock are available for decoding.
 	 */
-	XLogBeginInsert();
-	XLogRegisterData(&dummy_rec_data, 1);
-	wal_insert_ptr = XLogInsert(RM_XLOG_ID, XLOG_NOOP);
-	XLogFlush(wal_insert_ptr);
+	XLogFlush(GetXLogInsertRecPtr());
 	end_of_wal = GetFlushRecPtr(NULL);
 
 	/*
@@ -3186,10 +3129,7 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 	{
 		Oid			ind_old = lfirst_oid(lc);
 		Oid			ind_new = lfirst_oid(lc2);
-		Oid			mapped_tables[4];
-
-		/* Zero out possible results from swapped_relation_files */
-		memset(mapped_tables, 0, sizeof(mapped_tables));
+		Oid			mapped_tables[4] = {0};
 
 		swap_relation_files(ind_old, ind_new,
 							(old_table_oid == RelationRelationId),
@@ -3200,13 +3140,12 @@ rebuild_relation_finish_concurrent(Relation NewHeap, Relation OldHeap,
 							mapped_tables);
 
 #ifdef USE_ASSERT_CHECKING
-
 		/*
 		 * Concurrent processing is not supported for system relations, so
 		 * there should be no mapped tables.
 		 */
 		for (int i = 0; i < 4; i++)
-			Assert(mapped_tables[i] == 0);
+			Assert(!OidIsValid(mapped_tables[i]));
 #endif
 	}
 
@@ -3256,22 +3195,22 @@ build_new_indexes(Relation NewHeap, Relation OldHeap, List *OldIndexes)
 	pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
 								 PROGRESS_REPACK_PHASE_REBUILD_INDEX);
 
-	foreach_oid(ind_oid, OldIndexes)
+	foreach_oid(oldindex, OldIndexes)
 	{
-		Oid			ind_oid_new;
+		Oid			newindex;
 		char	   *newName;
 		Relation	ind;
 
-		ind = index_open(ind_oid, ShareUpdateExclusiveLock);
+		ind = index_open(oldindex, ShareUpdateExclusiveLock);
 
-		newName = ChooseRelationName(get_rel_name(ind_oid),
+		newName = ChooseRelationName(get_rel_name(oldindex),
 									 NULL,
 									 "repacknew",
 									 get_rel_namespace(ind->rd_index->indrelid),
 									 false);
-		ind_oid_new = index_create_copy(NewHeap, false, ind_oid,
-										ind->rd_rel->reltablespace, newName);
-		result = lappend_oid(result, ind_oid_new);
+		newindex = index_create_copy(NewHeap, false, oldindex,
+									 ind->rd_rel->reltablespace, newName);
+		result = lappend_oid(result, newindex);
 
 		index_close(ind, NoLock);
 	}
-- 
2.47.3


--2fxmamo6mu2qbxgv
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment;
	filename="v47-0005-Use-BulkInsertState-when-copying-data-to-the-new.patch"



^ permalink  raw  reply  [nested|flat] 197+ messages in thread