public inbox for [email protected]
help / color / mirror / Atom feedPartial aggregates pushdown
42+ messages / 13 participants
[nested] [flat]
* Partial aggregates pushdown
@ 2021-10-15 13:15 Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-10-15 13:15 UTC (permalink / raw)
To: pgsql-hackers
Hi.
One of the issues when we try to use sharding in PostgreSQL is absence
of partial aggregates pushdown.
I see several opportunities to alleviate this issue.
If we look at Citus, it implements aggregate, calculating internal state
of an arbitrary agregate function and exporting it as text. So we could
calculate internal states independently on all data sources and then
finalize it, which allows to compute arbitrary aggregate.
But, as mentioned in [1] thread, for some functions (like
count/max/min/sum) we can just push down them. It seems easy and covers
a lot of cases.
For now there are still issues - for example you can't handle functions
as avg() as we should somehow get its internal state or sum() variants,
which need aggserialfn/aggdeserialfn. Preliminary version is attached.
Is someone else working on the issue? Does suggested approach make
sense?
[1]
https://www.postgresql.org/message-id/flat/9998c3af9fdb5f7d62a6c7ad0fcd9142%40postgrespro.ru
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down.patch (22.0K, ../../[email protected]/2-0001-Partial-aggregates-push-down.patch)
download | inline diff:
From f2cf87a0ba1f4e4bf3f5f9e5b10b782c51717baf Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 59 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 215 +++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 29 ++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 31 ++-
4 files changed, 310 insertions(+), 24 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..0e4ac0ad255 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,53 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple proctup;
+ Form_pg_proc procform;
+ const char *proname;
+ bool ok = false;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ /* Can't process aggregates which require serialization/deserialization */
+ if (agg->aggtranstype == INTERNALOID)
+ return false;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ /* Only system aggregates are allowed */
+ if (procform->pronamespace != PG_CATALOG_NAMESPACE)
+ {
+ ReleaseSysCache(proctup);
+ return false;
+ }
+
+ /*
+ * Ensure that this partial aggregate can be evaluated directly. We check
+ * function name to avoid checking a dozen function variants.
+ */
+ proname = NameStr(procform->proname);
+
+ if (strcmp(proname, "min") == 0 ||
+ strcmp(proname, "max") == 0 ||
+ strcmp(proname, "count") == 0 ||
+ strcmp(proname, "sum") == 0)
+ ok = true;
+
+ ReleaseSysCache(proctup);
+
+ return ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 44c4367b8f9..89451e208e0 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,203 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 3000
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT sum(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: sum(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT sum(d) FROM pagg_tab;
+ sum
+-------
+ 58500
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..90e484bc640 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -517,7 +517,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +526,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -6083,7 +6084,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6098,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6342,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6357,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6386,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6403,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6425,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index e7b869f8cea..63c0c03da37 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,29 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*) FROM pagg_tab;
+SELECT max(a), count(*) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT sum(d) FROM pagg_tab;
+SELECT sum(d) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-10-15 14:56 ` Tomas Vondra <[email protected]>
2021-10-15 15:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 2 replies; 42+ messages in thread
From: Tomas Vondra @ 2021-10-15 14:56 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; pgsql-hackers
Hi Alexander,
On 10/15/21 15:15, Alexander Pyhalov wrote:
> Hi.
>
> One of the issues when we try to use sharding in PostgreSQL is absence
> of partial aggregates pushdown.
>
> I see several opportunities to alleviate this issue.
> If we look at Citus, it implements aggregate, calculating internal state
> of an arbitrary agregate function and exporting it as text. So we could
> calculate internal states independently on all data sources and then
> finalize it, which allows to compute arbitrary aggregate.
>
> But, as mentioned in [1] thread, for some functions (like
> count/max/min/sum) we can just push down them. It seems easy and covers
> a lot of cases.
> For now there are still issues - for example you can't handle functions
> as avg() as we should somehow get its internal state or sum() variants,
> which need aggserialfn/aggdeserialfn. Preliminary version is attached.
>
> Is someone else working on the issue? Does suggested approach make sense?
>
I think a couple people worked on this (or something similar/related) in
the past, but I don't recall any recent patches.
IMHO being able to push-down parts of an aggregation to other nodes is a
very desirable feature, that might result in huge improvements for some
analytical workloads.
As for the proposed approach, it's probably good enough for the first
version to restrict this to aggregates where the aggregate result is
sufficient, i.e. we don't need any new export/import procedures.
But it's very unlikely we'd want to restrict it the way the patch does
it, i.e. based on aggregate name. That's both fragile (people can create
new aggregates with such name) and against the PostgreSQL extensibility
(people may implement custom aggregates, but won't be able to benefit
from this just because of name).
So for v0 maybe, but I think there neeeds to be a way to relax this in
some way, for example we could add a new flag to pg_aggregate to mark
aggregates supporting this.
And then we should extend this for aggregates with more complex internal
states (e.g. avg), by supporting a function that "exports" the aggregate
state - similar to serial/deserial functions, but needs to be portable.
I think the trickiest thing here is rewriting the remote query to call
this export function, but maybe we could simply instruct the remote node
to use a different final function for the top-level node?
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
@ 2021-10-15 15:05 ` Alexander Pyhalov <[email protected]>
2021-10-15 15:26 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-10-15 15:05 UTC (permalink / raw)
To: Tomas Vondra <[email protected]>; +Cc: pgsql-hackers
Tomas Vondra писал 2021-10-15 17:56:
> Hi Alexander,
>
Hi.
> And then we should extend this for aggregates with more complex
> internal states (e.g. avg), by supporting a function that "exports"
> the aggregate state - similar to serial/deserial functions, but needs
> to be portable.
>
> I think the trickiest thing here is rewriting the remote query to call
> this export function, but maybe we could simply instruct the remote
> node to use a different final function for the top-level node?
>
>
If we have some special export function, how should we find out that
remote server supports this? Should it be server property or should it
somehow find out it while connecting to the server?
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-15 15:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-10-15 15:26 ` Tomas Vondra <[email protected]>
2021-10-15 19:31 ` Re: Partial aggregates pushdown Stephen Frost <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Tomas Vondra @ 2021-10-15 15:26 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: pgsql-hackers
On 10/15/21 17:05, Alexander Pyhalov wrote:
> Tomas Vondra писал 2021-10-15 17:56:
>> Hi Alexander,
>>
>
> Hi.
>
>> And then we should extend this for aggregates with more complex
>> internal states (e.g. avg), by supporting a function that "exports"
>> the aggregate state - similar to serial/deserial functions, but needs
>> to be portable.
>>
>> I think the trickiest thing here is rewriting the remote query to call
>> this export function, but maybe we could simply instruct the remote
>> node to use a different final function for the top-level node?
>>
>>
>
> If we have some special export function, how should we find out that
> remote server supports this? Should it be server property or should it
> somehow find out it while connecting to the server?
>
Good question. I guess there could be some initial negotiation based on
remote node version etc. And we could also disable this pushdown for
older server versions, etc.
But after that, I think we can treat this just like other definitions
between local/remote node - we'd assume they match (i.e. the remote
server has the export function), and then we'd get an error if it does
not. If you need to use remote nodes without an export function, you'd
have to disable the pushdown.
AFAICS this works both for case with explicit query rewrite (i.e. we
send SQL with calls to the export function) and implicit query rewrite
(where the remote node uses a different finalize function based on mode,
specified by GUC).
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-15 15:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 15:26 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
@ 2021-10-15 19:31 ` Stephen Frost <[email protected]>
2021-10-15 21:59 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Stephen Frost @ 2021-10-15 19:31 UTC (permalink / raw)
To: Tomas Vondra <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; pgsql-hackers
Greetings,
* Tomas Vondra ([email protected]) wrote:
> On 10/15/21 17:05, Alexander Pyhalov wrote:
> >Tomas Vondra писал 2021-10-15 17:56:
> >>And then we should extend this for aggregates with more complex
> >>internal states (e.g. avg), by supporting a function that "exports"
> >>the aggregate state - similar to serial/deserial functions, but needs
> >>to be portable.
> >>
> >>I think the trickiest thing here is rewriting the remote query to call
> >>this export function, but maybe we could simply instruct the remote
> >>node to use a different final function for the top-level node?
> >
> >If we have some special export function, how should we find out that
> >remote server supports this? Should it be server property or should it
> >somehow find out it while connecting to the server?
>
> Good question. I guess there could be some initial negotiation based on
> remote node version etc. And we could also disable this pushdown for older
> server versions, etc.
Yeah, I'd think we would just only support it on versions where we know
it's available. That doesn't seem terribly difficult.
> But after that, I think we can treat this just like other definitions
> between local/remote node - we'd assume they match (i.e. the remote server
> has the export function), and then we'd get an error if it does not. If you
> need to use remote nodes without an export function, you'd have to disable
> the pushdown.
>
> AFAICS this works both for case with explicit query rewrite (i.e. we send
> SQL with calls to the export function) and implicit query rewrite (where the
> remote node uses a different finalize function based on mode, specified by
> GUC).
Not quite sure where to drop this, but I've always figured we'd find a
way to use the existing PartialAgg / FinalizeAggregate bits which are
used for parallel query when it comes to pushing down to foreign servers
to perform aggregates. That also gives us how to serialize the results,
though we'd have to make sure that works across different
architectures.. I've not looked to see if that's the case today.
Then again, being able to transform an aggregate into a partial
aggregate that runs as an actual SQL query would mean we do partial
aggregate push-down against non-PG FDWs and that'd be pretty darn neat,
so maybe that's a better way to go, if we can figure out how.
(I mean, for avg it's pretty easy to just turn that into a SELECT that
grabs the sum and the count and use that.. other aggregates are more
complicated though and that doesn't work, maybe we need both?)
Thanks,
Stephen
Attachments:
[application/pgp-signature] signature.asc (819B, ../../[email protected]/2-signature.asc)
download
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-15 15:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 15:26 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-15 19:31 ` Re: Partial aggregates pushdown Stephen Frost <[email protected]>
@ 2021-10-15 21:59 ` Tomas Vondra <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Tomas Vondra @ 2021-10-15 21:59 UTC (permalink / raw)
To: Stephen Frost <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; pgsql-hackers
On 10/15/21 21:31, Stephen Frost wrote:
> Greetings,
>
> * Tomas Vondra ([email protected]) wrote:
>> On 10/15/21 17:05, Alexander Pyhalov wrote:
>>> Tomas Vondra писал 2021-10-15 17:56:
>>>> And then we should extend this for aggregates with more complex
>>>> internal states (e.g. avg), by supporting a function that "exports"
>>>> the aggregate state - similar to serial/deserial functions, but needs
>>>> to be portable.
>>>>
>>>> I think the trickiest thing here is rewriting the remote query to call
>>>> this export function, but maybe we could simply instruct the remote
>>>> node to use a different final function for the top-level node?
>>>
>>> If we have some special export function, how should we find out that
>>> remote server supports this? Should it be server property or should it
>>> somehow find out it while connecting to the server?
>>
>> Good question. I guess there could be some initial negotiation based on
>> remote node version etc. And we could also disable this pushdown for older
>> server versions, etc.
>
> Yeah, I'd think we would just only support it on versions where we know
> it's available. That doesn't seem terribly difficult.
>
Yeah.
But maybe Alexander was concerned about cases where the nodes disagree
on the aggregate definition, so one node might have the export function
and the other would not. E.g. the remote node may have older version of
an extension implementing the aggregate, without the export function
(although the server version supports it). I don't think we can do much
about that, it's just one of many issues that may be caused by
mismatching schemas.
I wonder if this might get more complex, though. Imagine for example a
partitioned table on node A with a FDW partition, pointing to a node B.
But on B, the object is partitioned again, with one partition placed on
C. So it's like
A -> partition on B -> partition on C
When planning on A, we can consider server version on B. But what if C
is an older version, not supporting the export function?
Bot sure if this makes any difference, though ... in the worst case it
will error out, and we should have a way to disable the feature on A.
>> But after that, I think we can treat this just like other definitions
>> between local/remote node - we'd assume they match (i.e. the remote server
>> has the export function), and then we'd get an error if it does not. If you
>> need to use remote nodes without an export function, you'd have to disable
>> the pushdown.
>>
>> AFAICS this works both for case with explicit query rewrite (i.e. we send
>> SQL with calls to the export function) and implicit query rewrite (where the
>> remote node uses a different finalize function based on mode, specified by
>> GUC).
>
> Not quite sure where to drop this, but I've always figured we'd find a
> way to use the existing PartialAgg / FinalizeAggregate bits which are
> used for parallel query when it comes to pushing down to foreign servers
> to perform aggregates. That also gives us how to serialize the results,
> though we'd have to make sure that works across different
> architectures.. I've not looked to see if that's the case today.
>
It sure is similar to what serial/deserial functions do for partial
aggs, but IIRC the functions were not designed to be portable. I think
we don't even require compatibility across minor releases, because we
only use this to copy data between workers running at the same time. Not
saying it can't be made to work, of course.
> Then again, being able to transform an aggregate into a partial
> aggregate that runs as an actual SQL query would mean we do partial
> aggregate push-down against non-PG FDWs and that'd be pretty darn neat,
> so maybe that's a better way to go, if we can figure out how.
>
> (I mean, for avg it's pretty easy to just turn that into a SELECT that
> grabs the sum and the count and use that.. other aggregates are more
> complicated though and that doesn't work, maybe we need both?)
>
Maybe, but that seems like a very different concept - transforming the
SQL so that it calculates different set of aggregates that we know can
be pushed down easily. But I don't recall any other practical example
beyond the AVG() -> SUM()/COUNT(). Well, VAR() can be translated into
SUM(X), SUM(X^2).
Another thing is how many users would actually benefit from this. I
mean, for this to matter you need partitioned table with partitions
placed on a non-PG FDW, right? Seems like a pretty niche use case.
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
@ 2021-10-19 06:56 ` Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-10-19 06:56 UTC (permalink / raw)
To: Tomas Vondra <[email protected]>; +Cc: pgsql-hackers
Hi.
Tomas Vondra писал 2021-10-15 17:56:
> As for the proposed approach, it's probably good enough for the first
> version to restrict this to aggregates where the aggregate result is
> sufficient, i.e. we don't need any new export/import procedures.
>
> But it's very unlikely we'd want to restrict it the way the patch does
> it, i.e. based on aggregate name. That's both fragile (people can
> create new aggregates with such name) and against the PostgreSQL
> extensibility (people may implement custom aggregates, but won't be
> able to benefit from this just because of name).
>
> So for v0 maybe, but I think there neeeds to be a way to relax this in
> some way, for example we could add a new flag to pg_aggregate to mark
> aggregates supporting this.
>
Updated patch to mark aggregates as pushdown-safe in pg_aggregates.
So far have no solution for aggregates with internal aggtranstype.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v02.patch (37.3K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v02.patch)
download | inline diff:
From 823a389caf003a21dd4c8e758f89d08ba89c5856 Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 45 +++-
.../postgres_fdw/expected/postgres_fdw.out | 215 +++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 29 ++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 31 ++-
src/backend/catalog/pg_aggregate.c | 4 +-
src/backend/commands/aggregatecmds.c | 6 +-
src/include/catalog/catversion.h | 2 +-
src/include/catalog/pg_aggregate.dat | 101 ++++----
src/include/catalog/pg_aggregate.h | 6 +-
9 files changed, 362 insertions(+), 77 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..cf6b2d9f066 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,39 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ /* Can't process aggregates which require serialization/deserialization */
+ if (agg->aggtranstype == INTERNALOID)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (aggform->aggpartialpushdownsafe != true)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ ReleaseSysCache(aggtup);
+
+ return true;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 44c4367b8f9..89451e208e0 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,203 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 3000
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT sum(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: sum(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL sum(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT sum(d) FROM pagg_tab;
+ sum
+-------
+ 58500
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..90e484bc640 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -517,7 +517,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +526,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -6083,7 +6084,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6098,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6342,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6357,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6386,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6403,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6425,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index e7b869f8cea..63c0c03da37 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,29 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*) FROM pagg_tab;
+SELECT max(a), count(*) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT sum(d) FROM pagg_tab;
+SELECT sum(d) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..8c91c87a85b 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -74,7 +74,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -676,6 +677,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..53976487fec 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -74,6 +74,7 @@ DefineAggregate(ParseState *pstate,
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -189,6 +190,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -471,7 +474,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 3253b8751b1..187c81b4857 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
*/
/* yyyymmddN */
-#define CATALOG_VERSION_NO 202109101
+#define CATALOG_VERSION_NO 202110191
#endif
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..e8df26e334c 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -60,22 +60,22 @@
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
aggcombinefn => 'float4pl', aggtranstype => 'float4' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +83,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..1138b64d478 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -91,6 +91,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -175,6 +178,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-10-19 13:25 ` Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Tomas Vondra @ 2021-10-19 13:25 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: pgsql-hackers
On 10/19/21 08:56, Alexander Pyhalov wrote:
> Hi.
>
> Tomas Vondra писал 2021-10-15 17:56:
>> As for the proposed approach, it's probably good enough for the first
>> version to restrict this to aggregates where the aggregate result is
>> sufficient, i.e. we don't need any new export/import procedures.
>>
>> But it's very unlikely we'd want to restrict it the way the patch does
>> it, i.e. based on aggregate name. That's both fragile (people can
>> create new aggregates with such name) and against the PostgreSQL
>> extensibility (people may implement custom aggregates, but won't be
>> able to benefit from this just because of name).
>>
>> So for v0 maybe, but I think there neeeds to be a way to relax this in
>> some way, for example we could add a new flag to pg_aggregate to mark
>> aggregates supporting this.
>>
>
> Updated patch to mark aggregates as pushdown-safe in pg_aggregates.
>
> So far have no solution for aggregates with internal aggtranstype.
Thanks. Please add it to the next CF, so that we don't lose track of it.
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
@ 2021-10-21 10:55 ` Alexander Pyhalov <[email protected]>
2021-11-01 09:47 ` Re: Partial aggregates pushdown Peter Eisentraut <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
0 siblings, 2 replies; 42+ messages in thread
From: Alexander Pyhalov @ 2021-10-21 10:55 UTC (permalink / raw)
To: Tomas Vondra <[email protected]>; +Cc: pgsql-hackers
Tomas Vondra писал 2021-10-19 16:25:
> On 10/19/21 08:56, Alexander Pyhalov wrote:
>> Hi.
>>
>> Tomas Vondra писал 2021-10-15 17:56:
>>> As for the proposed approach, it's probably good enough for the first
>>> version to restrict this to aggregates where the aggregate result is
>>> sufficient, i.e. we don't need any new export/import procedures.
>>>
>>> But it's very unlikely we'd want to restrict it the way the patch
>>> does
>>> it, i.e. based on aggregate name. That's both fragile (people can
>>> create new aggregates with such name) and against the PostgreSQL
>>> extensibility (people may implement custom aggregates, but won't be
>>> able to benefit from this just because of name).
>>>
>>> So for v0 maybe, but I think there neeeds to be a way to relax this
>>> in
>>> some way, for example we could add a new flag to pg_aggregate to mark
>>> aggregates supporting this.
>>>
>>
>> Updated patch to mark aggregates as pushdown-safe in pg_aggregates.
>>
>> So far have no solution for aggregates with internal aggtranstype.
Hi. Updated patch.
Now aggregates with internal states can be pushed down, if they are
marked as pushdown safe (this flag is set to true for min/max/sum),
have internal states and associated converters. Converters are called
locally, they transform aggregate result to serialized internal
representation.
As converters don't have access to internal aggregate state, partial
aggregates like avg() are still not pushable.
For now the overall logic is quite simple. We now also call
add_foreign_grouping_paths() for partial aggregation. In
foreign_expr_walker() we check if aggregate is pushable (which means
that it is simple, marked as pushable and if has 'internal' as
aggtranstype, has associated converter).
If it is pushable, we proceed as with usual aggregates (but forbid
having pushdown). During postgresGetForeignPlan() we produce list of
converters for aggregates. As converters has different input argument
type from their result (bytea), we have to generate alternative
metadata, which is used by make_tuple_from_result_row().
If make_tuple_from_result_row() encounters field with converter, it
calls converter and returns result. For now we expect converter to have
only one input and output argument. Existing converters just transform
input value to internal representation and return its serialized form.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v03.patch (57.8K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v03.patch)
download | inline diff:
From 52cd61fdb5cb5fceeacd832462468d8676f57ca6 Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 49 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 185 ++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 195 +++++++++++++++++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 27 ++-
src/backend/catalog/pg_aggregate.c | 28 ++-
src/backend/commands/aggregatecmds.c | 23 ++-
src/backend/utils/adt/numeric.c | 96 +++++++++
src/bin/pg_dump/pg_dump.c | 21 +-
src/include/catalog/catversion.h | 2 +-
src/include/catalog/pg_aggregate.dat | 106 +++++-----
src/include/catalog/pg_aggregate.h | 10 +-
src/include/catalog/pg_proc.dat | 6 +
src/test/regress/expected/oidjoins.out | 1 +
13 files changed, 666 insertions(+), 83 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..50ef1009b97 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,43 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ ok = false;
+
+ if (ok)
+ {
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (aggform->aggpartialpushdownsafe != true)
+ ok = false;
+
+ /*
+ * But if an aggregate requires serialization/deserialization, partial
+ * converter should be defined
+ */
+ if (ok && agg->aggtranstype == INTERNALOID)
+ ok = (aggform->aggpartialconverterfn != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+
+ return ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 44c4367b8f9..3e1b997875e 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,173 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*), sum(pagg_tab.d)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ max | count | sum
+-----+-------+-------
+ 29 | 3000 | 58500
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..b94bb2757a0 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_class.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/vacuum.h"
@@ -48,6 +50,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -80,7 +83,12 @@ enum FdwScanPrivateIndex
* String describing join i.e. names of relations being joined and types
* of join, added when the scan is join
*/
- FdwScanPrivateRelations
+ FdwScanPrivateRelations,
+
+ /*
+ * List of functions to convert partial aggregate result
+ */
+ FdwScanPrivateConvertors
};
/*
@@ -139,10 +147,13 @@ typedef struct PgFdwScanState
* for a foreign join scan. */
TupleDesc tupdesc; /* tuple descriptor of scan */
AttInMetadata *attinmeta; /* attribute datatype conversion metadata */
+ AttInMetadata *rcvd_attinmeta; /* metadata for received tuples, NULL if
+ * there's no converters */
/* extracted fdw_private data */
char *query; /* text of SELECT command */
List *retrieved_attrs; /* list of retrieved attribute numbers */
+ List *conv_list; /* list of converters */
/* for remote query execution */
PGconn *conn; /* connection for the scan */
@@ -510,6 +521,7 @@ static HeapTuple make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context);
static void conversion_error_callback(void *arg);
@@ -517,7 +529,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +538,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -541,6 +554,7 @@ static void merge_fdw_options(PgFdwRelationInfo *fpinfo,
const PgFdwRelationInfo *fpinfo_i);
static int get_batch_size_option(Relation rel);
+static List *build_conv_list(RelOptInfo *foreignrel);
/*
* Foreign-data wrapper handler function: return a struct with pointers
@@ -1233,6 +1247,7 @@ postgresGetForeignPlan(PlannerInfo *root,
List *local_exprs = NIL;
List *params_list = NIL;
List *fdw_scan_tlist = NIL;
+ List *fdw_conv_list = NIL;
List *fdw_recheck_quals = NIL;
List *retrieved_attrs;
StringInfoData sql;
@@ -1336,6 +1351,9 @@ postgresGetForeignPlan(PlannerInfo *root,
/* Build the list of columns to be fetched from the foreign server. */
fdw_scan_tlist = build_tlist_to_deparse(foreignrel);
+ /* Build the list of converters for partial aggregates. */
+ fdw_conv_list = build_conv_list(foreignrel);
+
/*
* Ensure that the outer plan produces a tuple whose descriptor
* matches our scan tuple slot. Also, remove the local conditions
@@ -1415,6 +1433,8 @@ postgresGetForeignPlan(PlannerInfo *root,
if (IS_JOIN_REL(foreignrel) || IS_UPPER_REL(foreignrel))
fdw_private = lappend(fdw_private,
makeString(fpinfo->relation_name));
+ if (IS_UPPER_REL(foreignrel))
+ fdw_private = lappend(fdw_private, fdw_conv_list);
/*
* Create the ForeignScan node for the given relation.
@@ -1433,6 +1453,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expecxted to return BYTEA, but real input type is likely different.
+ */
+static AttInMetadata *
+get_rcvd_attinmeta(TupleDesc tupdesc, List *conv_list)
+{
+ TupleDesc rcvd_tupdesc;
+
+ Assert(conv_list != NIL);
+
+ rcvd_tupdesc = CreateTupleDescCopy(tupdesc);
+ for (int i = 0; i < rcvd_tupdesc->natts; i++)
+ {
+ Oid converter = InvalidOid;
+ Form_pg_attribute att = TupleDescAttr(rcvd_tupdesc, i);
+
+ converter = list_nth_oid(conv_list, i);
+ if (converter != InvalidOid)
+ {
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(converter));
+
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", converter);
+
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ if (procform->pronargs != 1)
+ elog(ERROR, "converter %s is expected to have one argument", NameStr(procform->proname));
+
+ att->atttypid = procform->proargtypes.values[0];
+
+ ReleaseSysCache(proctup);
+ }
+ }
+
+ return TupleDescGetAttInMetadata(rcvd_tupdesc);
+}
+
/*
* Construct a tuple descriptor for the scan tuples handled by a foreign join.
*/
@@ -1545,6 +1607,12 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
FdwScanPrivateSelectSql));
fsstate->retrieved_attrs = (List *) list_nth(fsplan->fdw_private,
FdwScanPrivateRetrievedAttrs);
+
+ if (list_length(fsplan->fdw_private) > FdwScanPrivateConvertors)
+ fsstate->conv_list = (List *) list_nth(fsplan->fdw_private,
+ FdwScanPrivateConvertors);
+
+
fsstate->fetch_size = intVal(list_nth(fsplan->fdw_private,
FdwScanPrivateFetchSize));
@@ -1572,6 +1640,8 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
}
fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
+ if (fsstate->conv_list != NIL)
+ fsstate->rcvd_attinmeta = get_rcvd_attinmeta(fsstate->tupdesc, fsstate->conv_list);
/*
* Prepare for processing of parameters used in remote query, if any.
@@ -3815,13 +3885,21 @@ fetch_more_data(ForeignScanState *node)
for (i = 0; i < numrows; i++)
{
+ AttInMetadata *attinmeta;
+
Assert(IsA(node->ss.ps.plan, ForeignScan));
+ if (fsstate->rcvd_attinmeta)
+ attinmeta = fsstate->rcvd_attinmeta;
+ else
+ attinmeta = fsstate->attinmeta;
+
fsstate->tuples[i] =
make_tuple_from_result_row(res, i,
fsstate->rel,
- fsstate->attinmeta,
+ attinmeta,
fsstate->retrieved_attrs,
+ fsstate->conv_list,
node,
fsstate->temp_cxt);
}
@@ -4309,6 +4387,7 @@ store_returning_result(PgFdwModifyState *fmstate,
fmstate->rel,
fmstate->attinmeta,
fmstate->retrieved_attrs,
+ NIL,
NULL,
fmstate->temp_cxt);
@@ -4603,6 +4682,7 @@ get_returning_data(ForeignScanState *node)
dmstate->rel,
dmstate->attinmeta,
dmstate->retrieved_attrs,
+ NIL,
node,
dmstate->temp_cxt);
ExecStoreHeapTuple(newtup, slot, false);
@@ -5183,6 +5263,7 @@ analyze_row_processor(PGresult *res, int row, PgFdwAnalyzeState *astate)
astate->rel,
astate->attinmeta,
astate->retrieved_attrs,
+ NIL,
NULL,
astate->temp_cxt);
@@ -6083,7 +6164,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6178,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6422,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6437,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6466,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6483,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6505,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
@@ -7108,6 +7201,30 @@ complete_pending_request(AsyncRequest *areq)
TupIsNull(areq->result) ? 0.0 : 1.0);
}
+/*
+ * Interface to fmgr to call converters
+ */
+static Datum
+call_converter(Oid converter, Oid collation, Datum value, bool isnull, bool *res_isnull)
+{
+ LOCAL_FCINFO(fcinfo, 1);
+ FmgrInfo flinfo;
+ Datum result;
+
+ fmgr_info(converter, &flinfo);
+
+ InitFunctionCallInfoData(*fcinfo, &flinfo, 1, collation, NULL, NULL);
+
+ fcinfo->args[0].value = value;
+ fcinfo->args[0].isnull = isnull;
+
+ result = FunctionCallInvoke(fcinfo);
+
+ if (res_isnull)
+ *res_isnull = fcinfo->isnull;
+ return result;
+}
+
/*
* Create a tuple from the specified row of the PGresult.
*
@@ -7127,6 +7244,7 @@ make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context)
{
@@ -7178,6 +7296,7 @@ make_tuple_from_result_row(PGresult *res,
errcallback.previous = error_context_stack;
error_context_stack = &errcallback;
+
/*
* i indexes columns in the relation, j indexes columns in the PGresult.
*/
@@ -7209,6 +7328,20 @@ make_tuple_from_result_row(PGresult *res,
valstr,
attinmeta->attioparams[i - 1],
attinmeta->atttypmods[i - 1]);
+ if (conv_list != NIL)
+ {
+ Oid converter = list_nth_oid(conv_list, i - 1);
+
+ if (converter != InvalidOid)
+ {
+ Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+ bool res_isnull;
+
+ values[i - 1] = call_converter(converter, att->attcollation, values[i - 1], nulls[i - 1], &res_isnull);
+
+ nulls[i - 1] = res_isnull;
+ }
+ }
}
else if (i == SelfItemPointerAttributeNumber)
{
@@ -7472,3 +7605,45 @@ get_batch_size_option(Relation rel)
return batch_size;
}
+
+static List *
+build_conv_list(RelOptInfo *foreignrel)
+{
+ List *convlist = NIL;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) foreignrel->fdw_private;
+ ListCell *lc;
+
+ if (IS_UPPER_REL(foreignrel))
+ {
+ foreach(lc, fpinfo->grouped_tlist)
+ {
+ TargetEntry *tlentry = (TargetEntry *) lfirst(lc);
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
+ {
+ Aggref *agg = (Aggref *) tlentry->expr;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && agg->aggtranstype == INTERNALOID)
+ {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ converter_oid = aggform->aggpartialconverterfn;
+ Assert(converter_oid != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+ }
+ convlist = lappend_oid(convlist, converter_oid);
+ }
+ }
+
+ return convlist;
+}
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index e7b869f8cea..db3f12f30be 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,25 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..639ea4cf9a6 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -60,6 +60,7 @@ AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -74,7 +75,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -88,6 +90,7 @@ AggregateCreate(const char *aggName,
Oid combinefn = InvalidOid; /* can be omitted */
Oid serialfn = InvalidOid; /* can be omitted */
Oid deserialfn = InvalidOid; /* can be omitted */
+ Oid partialconverterfn = InvalidOid; /* can be omitted */
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
@@ -569,6 +572,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial converter, if present.
+ */
+ if (aggpartialconverterfnName)
+ {
+ fnArgs[0] = finaltype;
+
+ partialconverterfn = lookup_agg_function(aggpartialconverterfnName, 1,
+ fnArgs, InvalidOid,
+ &rettype);
+
+ if (rettype != BYTEAOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial serialization function %s is not %s",
+ NameListToString(aggserialfnName),
+ format_type_be(BYTEAOID))));
+
+ }
+
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -664,6 +688,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggcombinefn - 1] = ObjectIdGetDatum(combinefn);
values[Anum_pg_aggregate_aggserialfn - 1] = ObjectIdGetDatum(serialfn);
values[Anum_pg_aggregate_aggdeserialfn - 1] = ObjectIdGetDatum(deserialfn);
+ values[Anum_pg_aggregate_aggpartialconverterfn - 1] = ObjectIdGetDatum(partialconverterfn);
values[Anum_pg_aggregate_aggmtransfn - 1] = ObjectIdGetDatum(mtransfn);
values[Anum_pg_aggregate_aggminvtransfn - 1] = ObjectIdGetDatum(minvtransfn);
values[Anum_pg_aggregate_aggmfinalfn - 1] = ObjectIdGetDatum(mfinalfn);
@@ -676,6 +701,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..81cb17a119f 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -69,11 +69,13 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialconverterfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -142,6 +144,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialconverterfunc") == 0)
+ partialconverterfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -189,6 +193,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -372,6 +378,18 @@ DefineAggregate(ParseState *pstate,
errmsg("must specify both or neither of serialization and deserialization functions")));
}
+ if (partialconverterfuncName)
+ {
+ /*
+ * Serialization is only needed/allowed for transtype INTERNAL.
+ */
+ if (transTypeId != INTERNALOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
+ errmsg("serialization functions may be specified only when the aggregate transition data type is %s",
+ format_type_be(INTERNALOID))));
+ }
+
/*
* If a moving-aggregate transtype is specified, look that up. Same
* restrictions as for transtype.
@@ -457,6 +475,8 @@ DefineAggregate(ParseState *pstate,
combinefuncName, /* combine function name */
serialfuncName, /* serial function name */
deserialfuncName, /* deserial function name */
+ partialconverterfuncName, /* partial converter function
+ * name */
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
@@ -471,7 +491,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 1de744855f3..081481615c0 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -5770,6 +5770,52 @@ int8_avg_deserialize(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
+/*
+ * int8_sum_to_internal_serialize
+ * Convert int8 argument to serialized internal representation
+ */
+Datum
+int8_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ PolyNumAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = (Int128AggState *) palloc0(sizeof(Int128AggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
+ }
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+#ifdef HAVE_INT128
+ int128_to_numericvar(state->sumX, &tmp_var);
+#else
+ accum_sum_final(&state->sumX, &tmp_var);
+#endif
+ numericvar_serialize(&buf, &tmp_var);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Inverse transition functions to go with the above.
*/
@@ -5995,6 +6041,56 @@ numeric_sum(PG_FUNCTION_ARGS)
PG_RETURN_NUMERIC(result);
}
+/*
+ * numeric_sum_to_internal_serialize
+ * Convert numeric argument to serialized internal representation
+ */
+Datum
+numeric_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ NumericAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = makeNumericAggStateCurrentContext(false);
+
+ if (!PG_ARGISNULL(0))
+ do_numeric_accum(state, PG_GETARG_NUMERIC(0));
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+ accum_sum_final(&state->sumX, &tmp_var);
+ numericvar_serialize(&buf, &tmp_var);
+
+ /* maxScale */
+ pq_sendint32(&buf, state->maxScale);
+
+ /* maxScaleCount */
+ pq_sendint64(&buf, state->maxScaleCount);
+
+ /* NaNcount */
+ pq_sendint64(&buf, state->NaNcount);
+
+ /* pInfcount */
+ pq_sendint64(&buf, state->pInfcount);
+
+ /* nInfcount */
+ pq_sendint64(&buf, state->nInfcount);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Workhorse routine for the standard deviance and variance
* aggregates. 'state' is aggregate's transition state.
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index a485fb2d070..af3768f9e6e 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -14224,11 +14224,13 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggcombinefn;
const char *aggserialfn;
const char *aggdeserialfn;
+ const char *aggpartialconverterfn;
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
bool aggfinalextra;
bool aggmfinalextra;
+ bool aggpartialpushdownsafe;
char aggfinalmodify;
char aggmfinalmodify;
const char *aggsortop;
@@ -14313,11 +14315,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 150000)
+ appendPQExpBufferStr(query,
+ "aggpartialconverterfn,\n"
+ "aggpartialpushdownsafe\n");
+ else
+ appendPQExpBufferStr(query,
+ "'0' AS aggpartialconverterfn,\n"
+ "false as aggpartialpushdownsafe\n");
appendPQExpBuffer(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -14335,6 +14345,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ aggpartialconverterfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialconverterfn"));
+ aggpartialpushdownsafe = (PQgetvalue(res, 0, PQfnumber(res, "aggpartialpushdownsafe"))[0] == 't');
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -14429,6 +14441,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(aggpartialconverterfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALCONVERTERFUNC = %s", aggpartialconverterfn);
+ if (aggpartialpushdownsafe)
+ appendPQExpBufferStr(details, ",\n PARTIAL_PUSHDOWN_SAFE");
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 3253b8751b1..187c81b4857 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
*/
/* yyyymmddN */
-#define CATALOG_VERSION_NO 202109101
+#define CATALOG_VERSION_NO 202110191
#endif
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..61c4d812b34 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -56,26 +56,27 @@
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ aggpartialconverterfn => 'int8_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
- aggcombinefn => 'float4pl', aggtranstype => 'float4' },
+ aggcombinefn => 'float4pl', aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +84,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialconverterfn => 'numeric_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..2c63102ff31 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* function to convert aggregate result to bytea (0 if none) */
+ regproc aggpartialconverterfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -91,6 +94,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -161,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -175,6 +182,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index d068d6532ec..1cf23b15df0 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -4798,6 +4798,12 @@
{ oid => '2786', descr => 'aggregate serial function',
proname => 'int8_avg_serialize', prorettype => 'bytea',
proargtypes => 'internal', prosrc => 'int8_avg_serialize' },
+{ oid => '9630', descr => 'partial aggregate converter function',
+ proname => 'int8_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'int8_sum_to_internal_serialize' },
+{ oid => '9631', descr => 'partial aggregate converter function',
+ proname => 'numeric_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'numeric_sum_to_internal_serialize' },
{ oid => '2787', descr => 'aggregate deserial function',
proname => 'int8_avg_deserialize', prorettype => 'internal',
proargtypes => 'bytea internal', prosrc => 'int8_avg_deserialize' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 1461e947cdf..e1680054c30 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {aggpartialconverterfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-11-01 09:47 ` Peter Eisentraut <[email protected]>
2021-11-01 10:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Peter Eisentraut @ 2021-11-01 09:47 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; +Cc: pgsql-hackers
On 21.10.21 12:55, Alexander Pyhalov wrote:
> Now aggregates with internal states can be pushed down, if they are
> marked as pushdown safe (this flag is set to true for min/max/sum),
> have internal states and associated converters. Converters are called
> locally, they transform aggregate result to serialized internal
> representation.
> As converters don't have access to internal aggregate state, partial
> aggregates like avg() are still not pushable.
It seems to me that the system should be able to determine from the
existing aggregate catalog entry whether an aggregate can be pushed
down. For example, it could check aggtranstype != internal and similar.
A separate boolean flag should not be necessary. Or if it is, the
patch should provide some guidance about how an aggregate function
author should set it.
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 09:47 ` Re: Partial aggregates pushdown Peter Eisentraut <[email protected]>
@ 2021-11-01 10:30 ` Alexander Pyhalov <[email protected]>
2021-11-01 21:53 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-11-01 10:30 UTC (permalink / raw)
To: Peter Eisentraut <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers
Peter Eisentraut писал 2021-11-01 12:47:
> On 21.10.21 12:55, Alexander Pyhalov wrote:
>> Now aggregates with internal states can be pushed down, if they are
>> marked as pushdown safe (this flag is set to true for min/max/sum),
>> have internal states and associated converters. Converters are called
>> locally, they transform aggregate result to serialized internal
>> representation.
>> As converters don't have access to internal aggregate state, partial
>> aggregates like avg() are still not pushable.
>
> It seems to me that the system should be able to determine from the
> existing aggregate catalog entry whether an aggregate can be pushed
> down. For example, it could check aggtranstype != internal and
> similar. A separate boolean flag should not be necessary.
Hi.
I think we can't infer this property from existing flags. For example,
if I have avg() with bigint[] argtranstype, it doesn't mean we can push
down it. We couldn't also decide if partial aggregete is safe to push
down based on aggfinalfn presence (for example, it is defined for
sum(numeric), but we can push it down.
> Or if it
> is, the patch should provide some guidance about how an aggregate
> function author should set it.
Where should it be provided?
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 09:47 ` Re: Partial aggregates pushdown Peter Eisentraut <[email protected]>
2021-11-01 10:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-11-01 21:53 ` Ilya Gladyshev <[email protected]>
2021-11-01 23:49 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Ilya Gladyshev @ 2021-11-01 21:53 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Peter Eisentraut <[email protected]>
On 01.11.2021 13:30, Alexander Pyhalov wrote:
> Peter Eisentraut писал 2021-11-01 12:47:
>> On 21.10.21 12:55, Alexander Pyhalov wrote:
>>> Now aggregates with internal states can be pushed down, if they are
>>> marked as pushdown safe (this flag is set to true for min/max/sum),
>>> have internal states and associated converters. Converters are
>>> called locally, they transform aggregate result to serialized
>>> internal representation.
>>> As converters don't have access to internal aggregate state, partial
>>> aggregates like avg() are still not pushable.
>>
>> It seems to me that the system should be able to determine from the
>> existing aggregate catalog entry whether an aggregate can be pushed
>> down. For example, it could check aggtranstype != internal and
>> similar. A separate boolean flag should not be necessary.
>
> Hi.
> I think we can't infer this property from existing flags. For example,
> if I have avg() with bigint[] argtranstype, it doesn't mean we can
> push down it. We couldn't also decide if partial aggregete is safe to
> push down based on aggfinalfn presence (for example, it is defined for
> sum(numeric), but we can push it down.
I think one potential way to do it would be to allow pushing down
aggregates that EITHER have state of the same type as their return type,
OR have a conversion function that converts their return value to the
type of their state.
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 09:47 ` Re: Partial aggregates pushdown Peter Eisentraut <[email protected]>
2021-11-01 10:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:53 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
@ 2021-11-01 23:49 ` Tomas Vondra <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Tomas Vondra @ 2021-11-01 23:49 UTC (permalink / raw)
To: Ilya Gladyshev <[email protected]>; Alexander Pyhalov <[email protected]>; +Cc: pgsql-hackers; Peter Eisentraut <[email protected]>
On 11/1/21 22:53, Ilya Gladyshev wrote:
>
> On 01.11.2021 13:30, Alexander Pyhalov wrote:
>> Peter Eisentraut писал 2021-11-01 12:47:
>>> On 21.10.21 12:55, Alexander Pyhalov wrote:
>>>> Now aggregates with internal states can be pushed down, if they are
>>>> marked as pushdown safe (this flag is set to true for min/max/sum),
>>>> have internal states and associated converters. Converters are
>>>> called locally, they transform aggregate result to serialized
>>>> internal representation.
>>>> As converters don't have access to internal aggregate state, partial
>>>> aggregates like avg() are still not pushable.
>>>
>>> It seems to me that the system should be able to determine from the
>>> existing aggregate catalog entry whether an aggregate can be pushed
>>> down. For example, it could check aggtranstype != internal and
>>> similar. A separate boolean flag should not be necessary.
>>
>> Hi.
>> I think we can't infer this property from existing flags. For example,
>> if I have avg() with bigint[] argtranstype, it doesn't mean we can
>> push down it. We couldn't also decide if partial aggregete is safe to
>> push down based on aggfinalfn presence (for example, it is defined for
>> sum(numeric), but we can push it down.
>
> I think one potential way to do it would be to allow pushing down
> aggregates that EITHER have state of the same type as their return type,
> OR have a conversion function that converts their return value to the
> type of their state.
>
IMO just checking (aggtranstype == result type) entirely ignores the
issue of portability - we've never required the aggregate state to be
portable in any meaningful way (between architectures, minor/major
versions, ...) and it seems foolish to just start relying on it here.
Imagine for example an aggregate using bytea state, storing some complex
C struct in it. You can't just copy that between architectures.
It's a bit like why we don't simply copy data types to network, but pass
them through input/output or send/receive functions. The new flag is a
way to mark aggregates where this is safe, and I don't think we can do
away without it.
The more I think about this, the more I'm convinced the proper way to do
this would be adding export/import functions, similar to serial/deserial
functions, with the extra portability guarantees. And we'd need to do
that for all aggregates, not just those with (aggtranstype == internal).
I get it - the idea of the patch is that keeping the data types the same
makes it much simpler to pass the aggregate state (compared to having to
export/import it). But I'm not sure it's the right approach.
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-11-01 21:31 ` Ilya Gladyshev <[email protected]>
2021-11-01 21:57 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
1 sibling, 2 replies; 42+ messages in thread
From: Ilya Gladyshev @ 2021-11-01 21:31 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: pgsql-hackers
Hi,
On 21.10.2021 13:55, Alexander Pyhalov wrote:
> Hi. Updated patch.
> Now aggregates with internal states can be pushed down, if they are
> marked as pushdown safe (this flag is set to true for min/max/sum),
> have internal states and associated converters.
I don't quite understand why this is restricted only to aggregates that
have 'internal' state, I feel like that should be possible for any
aggregate that has a function to convert its final result back to
aggregate state to be pushed down. While I couldn't come up with a
useful example for this, except maybe for an aggregate whose aggfinalfn
is used purely for cosmetic purposes (e.g. format the result into a
string), I still feel that it is an unnecessary restriction.
A few minor review notes to the patch:
+static List *build_conv_list(RelOptInfo *foreignrel);
this should probably be up top among other declarations.
@@ -1433,6 +1453,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expecxted to return BYTEA, but real input type is likely
different.
+ */
typo in word "expec*x*ted".
@@ -139,10 +147,13 @@ typedef struct PgFdwScanState
* for a foreign join scan. */
TupleDesc tupdesc; /* tuple descriptor of scan */
AttInMetadata *attinmeta; /* attribute datatype conversion
metadata */
+ AttInMetadata *rcvd_attinmeta; /* metadata for received tuples,
NULL if
+ * there's no converters */
Looks like rcvd_attinmeta is redundant and you could use attinmeta for
conversion metadata.
The last thing - the patch needs to be rebased, it doesn't apply cleanly
on top of current master.
Thanks,
Ilya Gladyshev
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
@ 2021-11-01 21:57 ` Tomas Vondra <[email protected]>
1 sibling, 0 replies; 42+ messages in thread
From: Tomas Vondra @ 2021-11-01 21:57 UTC (permalink / raw)
To: Ilya Gladyshev <[email protected]>; Alexander Pyhalov <[email protected]>; +Cc: pgsql-hackers
On 11/1/21 22:31, Ilya Gladyshev wrote:
> Hi,
>
> On 21.10.2021 13:55, Alexander Pyhalov wrote:
>> Hi. Updated patch.
>> Now aggregates with internal states can be pushed down, if they are
>> marked as pushdown safe (this flag is set to true for min/max/sum),
>> have internal states and associated converters.
>
> I don't quite understand why this is restricted only to aggregates that
> have 'internal' state, I feel like that should be possible for any
> aggregate that has a function to convert its final result back to
> aggregate state to be pushed down. While I couldn't come up with a
> useful example for this, except maybe for an aggregate whose aggfinalfn
> is used purely for cosmetic purposes (e.g. format the result into a
> string), I still feel that it is an unnecessary restriction.
>
But it's *not* restricted to aggregates with internal state. The patch
merely requires aggregates with "internal" state to have an extra
"converter" function.
That being said, I don't think the approach used to deal with internal
state is the right one. AFAICS it simply runs the aggregate on the
remote node, finalizes is there, and then uses the converter function to
"expand" the partial result back into the internal state.
Unfortunately that only works for aggregates like "sum" where the result
is enough to rebuild the internal state, but it fails for anything more
complex (like "avg" or "var").
Earlier in this thread I mentioned this to serial/deserial functions,
and I think we need to do something like that for internal state. I.e.
we need to call the "serial" function on the remote node, and which
dumps the whole internal state, and then "deserial" on the local node.
regards
--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
@ 2021-11-02 09:12 ` Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-11-02 09:12 UTC (permalink / raw)
To: Ilya Gladyshev <[email protected]>; +Cc: pgsql-hackers
Hi.
Updated and rebased patch.
Ilya Gladyshev писал 2021-11-02 00:31:
> Hi,
> On 21.10.2021 13:55, Alexander Pyhalov wrote:
>
>> Hi. Updated patch.
>> Now aggregates with internal states can be pushed down, if they are
>> marked as pushdown safe (this flag is set to true for min/max/sum),
>> have internal states and associated converters.
>
> I don't quite understand why this is restricted only to aggregates
> that have 'internal' state, I feel like that should be possible for
> any aggregate that has a function to convert its final result back to
> aggregate state to be pushed down. While I couldn't come up with a
> useful example for this, except maybe for an aggregate whose
> aggfinalfn is used purely for cosmetic purposes (e.g. format the
> result into a string), I still feel that it is an unnecessary
> restriction.
>
I don't feel comfortable with it for the following reasons.
- Now partial converters translate aggregate result to serialized
internal representation.
In case when aggregate type is different from internal state,
we'd have to translate it to non-serialized internal representation,
so converters should skip serialization step. This seems like
introducing two
kind of converters.
- I don't see any system aggregates which would benefit from this.
However, it doesn't seem to be complex, and if it seems to be desirable,
it can be done.
For now introduced check that transtype matches aggregate type (or is
internal)
in partial_agg_ok().
> A few minor review notes to the patch:
>
> +static List *build_conv_list(RelOptInfo *foreignrel);
>
> this should probably be up top among other declarations.
>
Moved it upper.
> @@ -1433,6 +1453,48 @@ postgresGetForeignPlan(PlannerInfo *root,
> outer_plan);
> }
>
> +/*
> + * Generate attinmeta if there are some converters:
> + * they are expecxted to return BYTEA, but real input type is likely
> different.
> + */
>
> typo in word "expecxted".
Fixed.
>
> @@ -139,10 +147,13 @@ typedef struct PgFdwScanState
> * for a foreign join scan. */
> TupleDesc tupdesc; /* tuple descriptor of scan */
> AttInMetadata *attinmeta; /* attribute datatype conversion
> metadata */
> + AttInMetadata *rcvd_attinmeta; /* metadata for received
> tuples, NULL if
> + * there's no converters */
>
> Looks like rcvd_attinmeta is redundant and you could use attinmeta for
> conversion metadata.
Seems so, removed it.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v05.patch (58.0K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v05.patch)
download | inline diff:
From 71de85154f9ac78e99f4ce8bf9aa341fee748609 Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 57 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 185 ++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 196 +++++++++++++++++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 27 ++-
src/backend/catalog/pg_aggregate.c | 28 ++-
src/backend/commands/aggregatecmds.c | 23 +-
src/backend/utils/adt/numeric.c | 96 +++++++++
src/bin/pg_dump/pg_dump.c | 21 +-
src/include/catalog/catversion.h | 2 +-
src/include/catalog/pg_aggregate.dat | 106 +++++-----
src/include/catalog/pg_aggregate.h | 10 +-
src/include/catalog/pg_proc.dat | 6 +
src/test/regress/expected/oidjoins.out | 1 +
13 files changed, 674 insertions(+), 84 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..8cee12c1b2a 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,51 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (!aggform->aggpartialpushdownsafe)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /*
+ * If an aggregate requires serialization/deserialization, partial
+ * converter should be defined
+ */
+ if (agg->aggtranstype == INTERNALOID && aggform->aggpartialconverterfn == InvalidOid)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /* In this case we currently don't use converter */
+ if (agg->aggtranstype != INTERNALOID && get_func_rettype(agg->aggfnoid) != agg->aggtranstype)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return true;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index fd141a0fa5c..80c507783e6 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,173 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*), sum(pagg_tab.d)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ max | count | sum
+-----+-------+-------
+ 29 | 3000 | 58500
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..bad03c49e49 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_class.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/vacuum.h"
@@ -48,6 +50,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -80,7 +83,12 @@ enum FdwScanPrivateIndex
* String describing join i.e. names of relations being joined and types
* of join, added when the scan is join
*/
- FdwScanPrivateRelations
+ FdwScanPrivateRelations,
+
+ /*
+ * List of functions to convert partial aggregate result
+ */
+ FdwScanPrivateConvertors
};
/*
@@ -143,6 +151,7 @@ typedef struct PgFdwScanState
/* extracted fdw_private data */
char *query; /* text of SELECT command */
List *retrieved_attrs; /* list of retrieved attribute numbers */
+ List *conv_list; /* list of converters */
/* for remote query execution */
PGconn *conn; /* connection for the scan */
@@ -474,6 +483,7 @@ static void store_returning_result(PgFdwModifyState *fmstate,
TupleTableSlot *slot, PGresult *res);
static void finish_foreign_modify(PgFdwModifyState *fmstate);
static void deallocate_query(PgFdwModifyState *fmstate);
+static List *build_conv_list(RelOptInfo *foreignrel);
static List *build_remote_returning(Index rtindex, Relation rel,
List *returningList);
static void rebuild_fdw_scan_tlist(ForeignScan *fscan, List *tlist);
@@ -510,6 +520,7 @@ static HeapTuple make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context);
static void conversion_error_callback(void *arg);
@@ -517,7 +528,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +537,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -541,7 +553,6 @@ static void merge_fdw_options(PgFdwRelationInfo *fpinfo,
const PgFdwRelationInfo *fpinfo_i);
static int get_batch_size_option(Relation rel);
-
/*
* Foreign-data wrapper handler function: return a struct with pointers
* to my callback routines.
@@ -1233,6 +1244,7 @@ postgresGetForeignPlan(PlannerInfo *root,
List *local_exprs = NIL;
List *params_list = NIL;
List *fdw_scan_tlist = NIL;
+ List *fdw_conv_list = NIL;
List *fdw_recheck_quals = NIL;
List *retrieved_attrs;
StringInfoData sql;
@@ -1336,6 +1348,9 @@ postgresGetForeignPlan(PlannerInfo *root,
/* Build the list of columns to be fetched from the foreign server. */
fdw_scan_tlist = build_tlist_to_deparse(foreignrel);
+ /* Build the list of converters for partial aggregates. */
+ fdw_conv_list = build_conv_list(foreignrel);
+
/*
* Ensure that the outer plan produces a tuple whose descriptor
* matches our scan tuple slot. Also, remove the local conditions
@@ -1415,6 +1430,8 @@ postgresGetForeignPlan(PlannerInfo *root,
if (IS_JOIN_REL(foreignrel) || IS_UPPER_REL(foreignrel))
fdw_private = lappend(fdw_private,
makeString(fpinfo->relation_name));
+ if (IS_UPPER_REL(foreignrel))
+ fdw_private = lappend(fdw_private, fdw_conv_list);
/*
* Create the ForeignScan node for the given relation.
@@ -1433,6 +1450,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expected to return BYTEA, but real input type is likely different.
+ */
+static AttInMetadata *
+get_rcvd_attinmeta(TupleDesc tupdesc, List *conv_list)
+{
+ TupleDesc rcvd_tupdesc;
+
+ Assert(conv_list != NIL);
+
+ rcvd_tupdesc = CreateTupleDescCopy(tupdesc);
+ for (int i = 0; i < rcvd_tupdesc->natts; i++)
+ {
+ Oid converter = InvalidOid;
+ Form_pg_attribute att = TupleDescAttr(rcvd_tupdesc, i);
+
+ converter = list_nth_oid(conv_list, i);
+ if (converter != InvalidOid)
+ {
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(converter));
+
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", converter);
+
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ if (procform->pronargs != 1)
+ elog(ERROR, "converter %s is expected to have one argument", NameStr(procform->proname));
+
+ att->atttypid = procform->proargtypes.values[0];
+
+ ReleaseSysCache(proctup);
+ }
+ }
+
+ return TupleDescGetAttInMetadata(rcvd_tupdesc);
+}
+
/*
* Construct a tuple descriptor for the scan tuples handled by a foreign join.
*/
@@ -1545,6 +1604,12 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
FdwScanPrivateSelectSql));
fsstate->retrieved_attrs = (List *) list_nth(fsplan->fdw_private,
FdwScanPrivateRetrievedAttrs);
+
+ if (list_length(fsplan->fdw_private) > FdwScanPrivateConvertors)
+ fsstate->conv_list = (List *) list_nth(fsplan->fdw_private,
+ FdwScanPrivateConvertors);
+
+
fsstate->fetch_size = intVal(list_nth(fsplan->fdw_private,
FdwScanPrivateFetchSize));
@@ -1571,7 +1636,10 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
fsstate->tupdesc = get_tupdesc_for_join_scan_tuples(node);
}
- fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
+ if (fsstate->conv_list != NIL)
+ fsstate->attinmeta = get_rcvd_attinmeta(fsstate->tupdesc, fsstate->conv_list);
+ else
+ fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
/*
* Prepare for processing of parameters used in remote query, if any.
@@ -3822,6 +3890,7 @@ fetch_more_data(ForeignScanState *node)
fsstate->rel,
fsstate->attinmeta,
fsstate->retrieved_attrs,
+ fsstate->conv_list,
node,
fsstate->temp_cxt);
}
@@ -4309,6 +4378,7 @@ store_returning_result(PgFdwModifyState *fmstate,
fmstate->rel,
fmstate->attinmeta,
fmstate->retrieved_attrs,
+ NIL,
NULL,
fmstate->temp_cxt);
@@ -4603,6 +4673,7 @@ get_returning_data(ForeignScanState *node)
dmstate->rel,
dmstate->attinmeta,
dmstate->retrieved_attrs,
+ NIL,
node,
dmstate->temp_cxt);
ExecStoreHeapTuple(newtup, slot, false);
@@ -5183,6 +5254,7 @@ analyze_row_processor(PGresult *res, int row, PgFdwAnalyzeState *astate)
astate->rel,
astate->attinmeta,
astate->retrieved_attrs,
+ NIL,
NULL,
astate->temp_cxt);
@@ -6083,7 +6155,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6169,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6413,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6428,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6457,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6474,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6496,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
@@ -7108,6 +7192,30 @@ complete_pending_request(AsyncRequest *areq)
TupIsNull(areq->result) ? 0.0 : 1.0);
}
+/*
+ * Interface to fmgr to call converters
+ */
+static Datum
+call_converter(Oid converter, Oid collation, Datum value, bool isnull, bool *res_isnull)
+{
+ LOCAL_FCINFO(fcinfo, 1);
+ FmgrInfo flinfo;
+ Datum result;
+
+ fmgr_info(converter, &flinfo);
+
+ InitFunctionCallInfoData(*fcinfo, &flinfo, 1, collation, NULL, NULL);
+
+ fcinfo->args[0].value = value;
+ fcinfo->args[0].isnull = isnull;
+
+ result = FunctionCallInvoke(fcinfo);
+
+ if (res_isnull)
+ *res_isnull = fcinfo->isnull;
+ return result;
+}
+
/*
* Create a tuple from the specified row of the PGresult.
*
@@ -7127,6 +7235,7 @@ make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context)
{
@@ -7209,6 +7318,20 @@ make_tuple_from_result_row(PGresult *res,
valstr,
attinmeta->attioparams[i - 1],
attinmeta->atttypmods[i - 1]);
+ if (conv_list != NIL)
+ {
+ Oid converter = list_nth_oid(conv_list, i - 1);
+
+ if (converter != InvalidOid)
+ {
+ Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+ bool res_isnull;
+
+ values[i - 1] = call_converter(converter, att->attcollation, values[i - 1], nulls[i - 1], &res_isnull);
+
+ nulls[i - 1] = res_isnull;
+ }
+ }
}
else if (i == SelfItemPointerAttributeNumber)
{
@@ -7472,3 +7595,54 @@ get_batch_size_option(Relation rel)
return batch_size;
}
+
+/*
+ * For UPPER_REL build a list of converters, corresponding to tlist entries.
+ */
+static List *
+build_conv_list(RelOptInfo *foreignrel)
+{
+ List *conv_list = NIL;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) foreignrel->fdw_private;
+ ListCell *lc;
+
+ if (IS_UPPER_REL(foreignrel))
+ {
+ /* For UPPER_REL tlist matches grouped_tlist */
+ foreach(lc, fpinfo->grouped_tlist)
+ {
+ TargetEntry *tlentry = (TargetEntry *) lfirst(lc);
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
+ {
+ Aggref *agg = (Aggref *) tlentry->expr;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && agg->aggtranstype == INTERNALOID)
+ {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ converter_oid = aggform->aggpartialconverterfn;
+ Assert(converter_oid != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+ }
+
+ /*
+ * We append InvalidOid to conv_list to preserve one-to-one
+ * mapping between tlist and conv_list members.
+ */
+ conv_list = lappend_oid(conv_list, converter_oid);
+ }
+ }
+
+ return conv_list;
+}
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index 43c30d492da..a9771a9cc57 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,25 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..639ea4cf9a6 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -60,6 +60,7 @@ AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -74,7 +75,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -88,6 +90,7 @@ AggregateCreate(const char *aggName,
Oid combinefn = InvalidOid; /* can be omitted */
Oid serialfn = InvalidOid; /* can be omitted */
Oid deserialfn = InvalidOid; /* can be omitted */
+ Oid partialconverterfn = InvalidOid; /* can be omitted */
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
@@ -569,6 +572,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial converter, if present.
+ */
+ if (aggpartialconverterfnName)
+ {
+ fnArgs[0] = finaltype;
+
+ partialconverterfn = lookup_agg_function(aggpartialconverterfnName, 1,
+ fnArgs, InvalidOid,
+ &rettype);
+
+ if (rettype != BYTEAOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial serialization function %s is not %s",
+ NameListToString(aggserialfnName),
+ format_type_be(BYTEAOID))));
+
+ }
+
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -664,6 +688,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggcombinefn - 1] = ObjectIdGetDatum(combinefn);
values[Anum_pg_aggregate_aggserialfn - 1] = ObjectIdGetDatum(serialfn);
values[Anum_pg_aggregate_aggdeserialfn - 1] = ObjectIdGetDatum(deserialfn);
+ values[Anum_pg_aggregate_aggpartialconverterfn - 1] = ObjectIdGetDatum(partialconverterfn);
values[Anum_pg_aggregate_aggmtransfn - 1] = ObjectIdGetDatum(mtransfn);
values[Anum_pg_aggregate_aggminvtransfn - 1] = ObjectIdGetDatum(minvtransfn);
values[Anum_pg_aggregate_aggmfinalfn - 1] = ObjectIdGetDatum(mfinalfn);
@@ -676,6 +701,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..3b7c1597315 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -69,11 +69,13 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialconverterfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -142,6 +144,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialconverterfunc") == 0)
+ partialconverterfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -189,6 +193,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -372,6 +378,18 @@ DefineAggregate(ParseState *pstate,
errmsg("must specify both or neither of serialization and deserialization functions")));
}
+ if (partialconverterfuncName)
+ {
+ /*
+ * Converter is only needed/allowed for transtype INTERNAL.
+ */
+ if (transTypeId != INTERNALOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
+ errmsg("partial converters may be specified only when the aggregate transition data type is %s",
+ format_type_be(INTERNALOID))));
+ }
+
/*
* If a moving-aggregate transtype is specified, look that up. Same
* restrictions as for transtype.
@@ -457,6 +475,8 @@ DefineAggregate(ParseState *pstate,
combinefuncName, /* combine function name */
serialfuncName, /* serial function name */
deserialfuncName, /* deserial function name */
+ partialconverterfuncName, /* partial converter function
+ * name */
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
@@ -471,7 +491,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 1de744855f3..e26e3e2de41 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -5770,6 +5770,52 @@ int8_avg_deserialize(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
+/*
+ * int8_sum_to_internal_serialize
+ * Convert int8 argument to serialized internal representation
+ */
+Datum
+int8_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ PolyNumAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = (Int128AggState *) palloc0(sizeof(Int128AggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
+ }
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+#ifdef HAVE_INT128
+ int128_to_numericvar(state->sumX, &tmp_var);
+#else
+ accum_sum_final(&state->sumX, &tmp_var);
+#endif
+ numericvar_serialize(&buf, &tmp_var);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Inverse transition functions to go with the above.
*/
@@ -5995,6 +6041,56 @@ numeric_sum(PG_FUNCTION_ARGS)
PG_RETURN_NUMERIC(result);
}
+/*
+ * numeric_sum_to_internal_serialize
+ * Convert numeric argument to serialized internal representation
+ */
+Datum
+numeric_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ NumericAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = makeNumericAggStateCurrentContext(false);
+
+ if (!PG_ARGISNULL(0))
+ do_numeric_accum(state, PG_GETARG_NUMERIC(0));
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+ accum_sum_final(&state->sumX, &tmp_var);
+ numericvar_serialize(&buf, &tmp_var);
+
+ /* maxScale */
+ pq_sendint32(&buf, state->maxScale);
+
+ /* maxScaleCount */
+ pq_sendint64(&buf, state->maxScaleCount);
+
+ /* NaNcount */
+ pq_sendint64(&buf, state->NaNcount);
+
+ /* pInfcount */
+ pq_sendint64(&buf, state->pInfcount);
+
+ /* nInfcount */
+ pq_sendint64(&buf, state->nInfcount);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Workhorse routine for the standard deviance and variance
* aggregates. 'state' is aggregate's transition state.
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index b9635a95b6f..5d9509ace0e 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -14076,11 +14076,13 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggcombinefn;
const char *aggserialfn;
const char *aggdeserialfn;
+ const char *aggpartialconverterfn;
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
bool aggfinalextra;
bool aggmfinalextra;
+ bool aggpartialpushdownsafe;
char aggfinalmodify;
char aggmfinalmodify;
const char *aggsortop;
@@ -14165,11 +14167,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 150000)
+ appendPQExpBufferStr(query,
+ "aggpartialconverterfn,\n"
+ "aggpartialpushdownsafe\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS aggpartialconverterfn,\n"
+ "false as aggpartialpushdownsafe\n");
appendPQExpBuffer(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -14187,6 +14197,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ aggpartialconverterfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialconverterfn"));
+ aggpartialpushdownsafe = (PQgetvalue(res, 0, PQfnumber(res, "aggpartialpushdownsafe"))[0] == 't');
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -14281,6 +14293,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(aggpartialconverterfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALCONVERTERFUNC = %s", aggpartialconverterfn);
+ if (aggpartialpushdownsafe)
+ appendPQExpBufferStr(details, ",\n PARTIAL_PUSHDOWN_SAFE");
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 9faf017457a..35c399e77dd 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
*/
/* yyyymmddN */
-#define CATALOG_VERSION_NO 202110272
+#define CATALOG_VERSION_NO 202111021
#endif
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..61c4d812b34 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -56,26 +56,27 @@
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ aggpartialconverterfn => 'int8_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
- aggcombinefn => 'float4pl', aggtranstype => 'float4' },
+ aggcombinefn => 'float4pl', aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +84,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialconverterfn => 'numeric_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..2c63102ff31 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* function to convert aggregate result to bytea (0 if none) */
+ regproc aggpartialconverterfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -91,6 +94,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -161,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -175,6 +182,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index d068d6532ec..1cf23b15df0 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -4798,6 +4798,12 @@
{ oid => '2786', descr => 'aggregate serial function',
proname => 'int8_avg_serialize', prorettype => 'bytea',
proargtypes => 'internal', prosrc => 'int8_avg_serialize' },
+{ oid => '9630', descr => 'partial aggregate converter function',
+ proname => 'int8_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'int8_sum_to_internal_serialize' },
+{ oid => '9631', descr => 'partial aggregate converter function',
+ proname => 'numeric_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'numeric_sum_to_internal_serialize' },
{ oid => '2787', descr => 'aggregate deserial function',
proname => 'int8_avg_deserialize', prorettype => 'internal',
proargtypes => 'bytea internal', prosrc => 'int8_avg_deserialize' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be3..864151bbca3 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {aggpartialconverterfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-11-03 13:45 ` Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Daniel Gustafsson @ 2021-11-03 13:45 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Ilya Gladyshev <[email protected]>; pgsql-hackers
> On 2 Nov 2021, at 10:12, Alexander Pyhalov <[email protected]> wrote:
> Updated and rebased patch.
+ state = (Int128AggState *) palloc0(sizeof(Int128AggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
This fails on non-INT128 platforms as state cannot be cast to Int128AggState
outside of HAVE_INT128; it's not defined there. This needs to be a
PolyNumAggState no?
--
Daniel Gustafsson https://vmware.com/
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
@ 2021-11-03 14:50 ` Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-11-03 14:50 UTC (permalink / raw)
To: Daniel Gustafsson <[email protected]>; +Cc: Ilya Gladyshev <[email protected]>; pgsql-hackers
Daniel Gustafsson писал 2021-11-03 16:45:
>> On 2 Nov 2021, at 10:12, Alexander Pyhalov <[email protected]>
>> wrote:
>
>> Updated and rebased patch.
>
> + state = (Int128AggState *) palloc0(sizeof(Int128AggState));
> + state->calcSumX2 = false;
> +
> + if (!PG_ARGISNULL(0))
> + {
> +#ifdef HAVE_INT128
> + do_int128_accum(state, (int128) PG_GETARG_INT64(0));
> +#else
> + do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
> +#endif
>
> This fails on non-INT128 platforms as state cannot be cast to
> Int128AggState
> outside of HAVE_INT128; it's not defined there. This needs to be a
> PolyNumAggState no?
Hi.
Thank you for noticing this. It's indeed fails with
pgac_cv__128bit_int=no.
Updated patch.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v06.patch (58.0K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v06.patch)
download | inline diff:
From f72a3d52a2b85ad9ea5f61f8ff5c46cb50ae3ec8 Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 57 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 185 ++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 196 +++++++++++++++++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 27 ++-
src/backend/catalog/pg_aggregate.c | 28 ++-
src/backend/commands/aggregatecmds.c | 23 +-
src/backend/utils/adt/numeric.c | 96 +++++++++
src/bin/pg_dump/pg_dump.c | 21 +-
src/include/catalog/catversion.h | 2 +-
src/include/catalog/pg_aggregate.dat | 106 +++++-----
src/include/catalog/pg_aggregate.h | 10 +-
src/include/catalog/pg_proc.dat | 6 +
src/test/regress/expected/oidjoins.out | 1 +
13 files changed, 674 insertions(+), 84 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..8cee12c1b2a 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,51 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (!aggform->aggpartialpushdownsafe)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /*
+ * If an aggregate requires serialization/deserialization, partial
+ * converter should be defined
+ */
+ if (agg->aggtranstype == INTERNALOID && aggform->aggpartialconverterfn == InvalidOid)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /* In this case we currently don't use converter */
+ if (agg->aggtranstype != INTERNALOID && get_func_rettype(agg->aggfnoid) != agg->aggtranstype)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return true;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index fd141a0fa5c..80c507783e6 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,173 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*), sum(pagg_tab.d)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ max | count | sum
+-----+-------+-------
+ 29 | 3000 | 58500
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..bad03c49e49 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_class.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/vacuum.h"
@@ -48,6 +50,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -80,7 +83,12 @@ enum FdwScanPrivateIndex
* String describing join i.e. names of relations being joined and types
* of join, added when the scan is join
*/
- FdwScanPrivateRelations
+ FdwScanPrivateRelations,
+
+ /*
+ * List of functions to convert partial aggregate result
+ */
+ FdwScanPrivateConvertors
};
/*
@@ -143,6 +151,7 @@ typedef struct PgFdwScanState
/* extracted fdw_private data */
char *query; /* text of SELECT command */
List *retrieved_attrs; /* list of retrieved attribute numbers */
+ List *conv_list; /* list of converters */
/* for remote query execution */
PGconn *conn; /* connection for the scan */
@@ -474,6 +483,7 @@ static void store_returning_result(PgFdwModifyState *fmstate,
TupleTableSlot *slot, PGresult *res);
static void finish_foreign_modify(PgFdwModifyState *fmstate);
static void deallocate_query(PgFdwModifyState *fmstate);
+static List *build_conv_list(RelOptInfo *foreignrel);
static List *build_remote_returning(Index rtindex, Relation rel,
List *returningList);
static void rebuild_fdw_scan_tlist(ForeignScan *fscan, List *tlist);
@@ -510,6 +520,7 @@ static HeapTuple make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context);
static void conversion_error_callback(void *arg);
@@ -517,7 +528,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +537,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -541,7 +553,6 @@ static void merge_fdw_options(PgFdwRelationInfo *fpinfo,
const PgFdwRelationInfo *fpinfo_i);
static int get_batch_size_option(Relation rel);
-
/*
* Foreign-data wrapper handler function: return a struct with pointers
* to my callback routines.
@@ -1233,6 +1244,7 @@ postgresGetForeignPlan(PlannerInfo *root,
List *local_exprs = NIL;
List *params_list = NIL;
List *fdw_scan_tlist = NIL;
+ List *fdw_conv_list = NIL;
List *fdw_recheck_quals = NIL;
List *retrieved_attrs;
StringInfoData sql;
@@ -1336,6 +1348,9 @@ postgresGetForeignPlan(PlannerInfo *root,
/* Build the list of columns to be fetched from the foreign server. */
fdw_scan_tlist = build_tlist_to_deparse(foreignrel);
+ /* Build the list of converters for partial aggregates. */
+ fdw_conv_list = build_conv_list(foreignrel);
+
/*
* Ensure that the outer plan produces a tuple whose descriptor
* matches our scan tuple slot. Also, remove the local conditions
@@ -1415,6 +1430,8 @@ postgresGetForeignPlan(PlannerInfo *root,
if (IS_JOIN_REL(foreignrel) || IS_UPPER_REL(foreignrel))
fdw_private = lappend(fdw_private,
makeString(fpinfo->relation_name));
+ if (IS_UPPER_REL(foreignrel))
+ fdw_private = lappend(fdw_private, fdw_conv_list);
/*
* Create the ForeignScan node for the given relation.
@@ -1433,6 +1450,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expected to return BYTEA, but real input type is likely different.
+ */
+static AttInMetadata *
+get_rcvd_attinmeta(TupleDesc tupdesc, List *conv_list)
+{
+ TupleDesc rcvd_tupdesc;
+
+ Assert(conv_list != NIL);
+
+ rcvd_tupdesc = CreateTupleDescCopy(tupdesc);
+ for (int i = 0; i < rcvd_tupdesc->natts; i++)
+ {
+ Oid converter = InvalidOid;
+ Form_pg_attribute att = TupleDescAttr(rcvd_tupdesc, i);
+
+ converter = list_nth_oid(conv_list, i);
+ if (converter != InvalidOid)
+ {
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(converter));
+
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", converter);
+
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ if (procform->pronargs != 1)
+ elog(ERROR, "converter %s is expected to have one argument", NameStr(procform->proname));
+
+ att->atttypid = procform->proargtypes.values[0];
+
+ ReleaseSysCache(proctup);
+ }
+ }
+
+ return TupleDescGetAttInMetadata(rcvd_tupdesc);
+}
+
/*
* Construct a tuple descriptor for the scan tuples handled by a foreign join.
*/
@@ -1545,6 +1604,12 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
FdwScanPrivateSelectSql));
fsstate->retrieved_attrs = (List *) list_nth(fsplan->fdw_private,
FdwScanPrivateRetrievedAttrs);
+
+ if (list_length(fsplan->fdw_private) > FdwScanPrivateConvertors)
+ fsstate->conv_list = (List *) list_nth(fsplan->fdw_private,
+ FdwScanPrivateConvertors);
+
+
fsstate->fetch_size = intVal(list_nth(fsplan->fdw_private,
FdwScanPrivateFetchSize));
@@ -1571,7 +1636,10 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
fsstate->tupdesc = get_tupdesc_for_join_scan_tuples(node);
}
- fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
+ if (fsstate->conv_list != NIL)
+ fsstate->attinmeta = get_rcvd_attinmeta(fsstate->tupdesc, fsstate->conv_list);
+ else
+ fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
/*
* Prepare for processing of parameters used in remote query, if any.
@@ -3822,6 +3890,7 @@ fetch_more_data(ForeignScanState *node)
fsstate->rel,
fsstate->attinmeta,
fsstate->retrieved_attrs,
+ fsstate->conv_list,
node,
fsstate->temp_cxt);
}
@@ -4309,6 +4378,7 @@ store_returning_result(PgFdwModifyState *fmstate,
fmstate->rel,
fmstate->attinmeta,
fmstate->retrieved_attrs,
+ NIL,
NULL,
fmstate->temp_cxt);
@@ -4603,6 +4673,7 @@ get_returning_data(ForeignScanState *node)
dmstate->rel,
dmstate->attinmeta,
dmstate->retrieved_attrs,
+ NIL,
node,
dmstate->temp_cxt);
ExecStoreHeapTuple(newtup, slot, false);
@@ -5183,6 +5254,7 @@ analyze_row_processor(PGresult *res, int row, PgFdwAnalyzeState *astate)
astate->rel,
astate->attinmeta,
astate->retrieved_attrs,
+ NIL,
NULL,
astate->temp_cxt);
@@ -6083,7 +6155,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6169,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6413,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6428,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6457,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6474,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6496,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
@@ -7108,6 +7192,30 @@ complete_pending_request(AsyncRequest *areq)
TupIsNull(areq->result) ? 0.0 : 1.0);
}
+/*
+ * Interface to fmgr to call converters
+ */
+static Datum
+call_converter(Oid converter, Oid collation, Datum value, bool isnull, bool *res_isnull)
+{
+ LOCAL_FCINFO(fcinfo, 1);
+ FmgrInfo flinfo;
+ Datum result;
+
+ fmgr_info(converter, &flinfo);
+
+ InitFunctionCallInfoData(*fcinfo, &flinfo, 1, collation, NULL, NULL);
+
+ fcinfo->args[0].value = value;
+ fcinfo->args[0].isnull = isnull;
+
+ result = FunctionCallInvoke(fcinfo);
+
+ if (res_isnull)
+ *res_isnull = fcinfo->isnull;
+ return result;
+}
+
/*
* Create a tuple from the specified row of the PGresult.
*
@@ -7127,6 +7235,7 @@ make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context)
{
@@ -7209,6 +7318,20 @@ make_tuple_from_result_row(PGresult *res,
valstr,
attinmeta->attioparams[i - 1],
attinmeta->atttypmods[i - 1]);
+ if (conv_list != NIL)
+ {
+ Oid converter = list_nth_oid(conv_list, i - 1);
+
+ if (converter != InvalidOid)
+ {
+ Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+ bool res_isnull;
+
+ values[i - 1] = call_converter(converter, att->attcollation, values[i - 1], nulls[i - 1], &res_isnull);
+
+ nulls[i - 1] = res_isnull;
+ }
+ }
}
else if (i == SelfItemPointerAttributeNumber)
{
@@ -7472,3 +7595,54 @@ get_batch_size_option(Relation rel)
return batch_size;
}
+
+/*
+ * For UPPER_REL build a list of converters, corresponding to tlist entries.
+ */
+static List *
+build_conv_list(RelOptInfo *foreignrel)
+{
+ List *conv_list = NIL;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) foreignrel->fdw_private;
+ ListCell *lc;
+
+ if (IS_UPPER_REL(foreignrel))
+ {
+ /* For UPPER_REL tlist matches grouped_tlist */
+ foreach(lc, fpinfo->grouped_tlist)
+ {
+ TargetEntry *tlentry = (TargetEntry *) lfirst(lc);
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
+ {
+ Aggref *agg = (Aggref *) tlentry->expr;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && agg->aggtranstype == INTERNALOID)
+ {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ converter_oid = aggform->aggpartialconverterfn;
+ Assert(converter_oid != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+ }
+
+ /*
+ * We append InvalidOid to conv_list to preserve one-to-one
+ * mapping between tlist and conv_list members.
+ */
+ conv_list = lappend_oid(conv_list, converter_oid);
+ }
+ }
+
+ return conv_list;
+}
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index 43c30d492da..a9771a9cc57 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,25 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..639ea4cf9a6 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -60,6 +60,7 @@ AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -74,7 +75,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -88,6 +90,7 @@ AggregateCreate(const char *aggName,
Oid combinefn = InvalidOid; /* can be omitted */
Oid serialfn = InvalidOid; /* can be omitted */
Oid deserialfn = InvalidOid; /* can be omitted */
+ Oid partialconverterfn = InvalidOid; /* can be omitted */
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
@@ -569,6 +572,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial converter, if present.
+ */
+ if (aggpartialconverterfnName)
+ {
+ fnArgs[0] = finaltype;
+
+ partialconverterfn = lookup_agg_function(aggpartialconverterfnName, 1,
+ fnArgs, InvalidOid,
+ &rettype);
+
+ if (rettype != BYTEAOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial serialization function %s is not %s",
+ NameListToString(aggserialfnName),
+ format_type_be(BYTEAOID))));
+
+ }
+
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -664,6 +688,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggcombinefn - 1] = ObjectIdGetDatum(combinefn);
values[Anum_pg_aggregate_aggserialfn - 1] = ObjectIdGetDatum(serialfn);
values[Anum_pg_aggregate_aggdeserialfn - 1] = ObjectIdGetDatum(deserialfn);
+ values[Anum_pg_aggregate_aggpartialconverterfn - 1] = ObjectIdGetDatum(partialconverterfn);
values[Anum_pg_aggregate_aggmtransfn - 1] = ObjectIdGetDatum(mtransfn);
values[Anum_pg_aggregate_aggminvtransfn - 1] = ObjectIdGetDatum(minvtransfn);
values[Anum_pg_aggregate_aggmfinalfn - 1] = ObjectIdGetDatum(mfinalfn);
@@ -676,6 +701,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..3b7c1597315 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -69,11 +69,13 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialconverterfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -142,6 +144,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialconverterfunc") == 0)
+ partialconverterfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -189,6 +193,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -372,6 +378,18 @@ DefineAggregate(ParseState *pstate,
errmsg("must specify both or neither of serialization and deserialization functions")));
}
+ if (partialconverterfuncName)
+ {
+ /*
+ * Converter is only needed/allowed for transtype INTERNAL.
+ */
+ if (transTypeId != INTERNALOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
+ errmsg("partial converters may be specified only when the aggregate transition data type is %s",
+ format_type_be(INTERNALOID))));
+ }
+
/*
* If a moving-aggregate transtype is specified, look that up. Same
* restrictions as for transtype.
@@ -457,6 +475,8 @@ DefineAggregate(ParseState *pstate,
combinefuncName, /* combine function name */
serialfuncName, /* serial function name */
deserialfuncName, /* deserial function name */
+ partialconverterfuncName, /* partial converter function
+ * name */
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
@@ -471,7 +491,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 1de744855f3..ac900867e87 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -5770,6 +5770,52 @@ int8_avg_deserialize(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
+/*
+ * int8_sum_to_internal_serialize
+ * Convert int8 argument to serialized internal representation
+ */
+Datum
+int8_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ PolyNumAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = (PolyNumAggState *) palloc0(sizeof(PolyNumAggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
+ }
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+#ifdef HAVE_INT128
+ int128_to_numericvar(state->sumX, &tmp_var);
+#else
+ accum_sum_final(&state->sumX, &tmp_var);
+#endif
+ numericvar_serialize(&buf, &tmp_var);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Inverse transition functions to go with the above.
*/
@@ -5995,6 +6041,56 @@ numeric_sum(PG_FUNCTION_ARGS)
PG_RETURN_NUMERIC(result);
}
+/*
+ * numeric_sum_to_internal_serialize
+ * Convert numeric argument to serialized internal representation
+ */
+Datum
+numeric_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ NumericAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = makeNumericAggStateCurrentContext(false);
+
+ if (!PG_ARGISNULL(0))
+ do_numeric_accum(state, PG_GETARG_NUMERIC(0));
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+ accum_sum_final(&state->sumX, &tmp_var);
+ numericvar_serialize(&buf, &tmp_var);
+
+ /* maxScale */
+ pq_sendint32(&buf, state->maxScale);
+
+ /* maxScaleCount */
+ pq_sendint64(&buf, state->maxScaleCount);
+
+ /* NaNcount */
+ pq_sendint64(&buf, state->NaNcount);
+
+ /* pInfcount */
+ pq_sendint64(&buf, state->pInfcount);
+
+ /* nInfcount */
+ pq_sendint64(&buf, state->nInfcount);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Workhorse routine for the standard deviance and variance
* aggregates. 'state' is aggregate's transition state.
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index b9635a95b6f..5d9509ace0e 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -14076,11 +14076,13 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggcombinefn;
const char *aggserialfn;
const char *aggdeserialfn;
+ const char *aggpartialconverterfn;
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
bool aggfinalextra;
bool aggmfinalextra;
+ bool aggpartialpushdownsafe;
char aggfinalmodify;
char aggmfinalmodify;
const char *aggsortop;
@@ -14165,11 +14167,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 150000)
+ appendPQExpBufferStr(query,
+ "aggpartialconverterfn,\n"
+ "aggpartialpushdownsafe\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS aggpartialconverterfn,\n"
+ "false as aggpartialpushdownsafe\n");
appendPQExpBuffer(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -14187,6 +14197,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ aggpartialconverterfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialconverterfn"));
+ aggpartialpushdownsafe = (PQgetvalue(res, 0, PQfnumber(res, "aggpartialpushdownsafe"))[0] == 't');
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -14281,6 +14293,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(aggpartialconverterfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALCONVERTERFUNC = %s", aggpartialconverterfn);
+ if (aggpartialpushdownsafe)
+ appendPQExpBufferStr(details, ",\n PARTIAL_PUSHDOWN_SAFE");
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 9faf017457a..35c399e77dd 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
*/
/* yyyymmddN */
-#define CATALOG_VERSION_NO 202110272
+#define CATALOG_VERSION_NO 202111021
#endif
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..61c4d812b34 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -56,26 +56,27 @@
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ aggpartialconverterfn => 'int8_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
- aggcombinefn => 'float4pl', aggtranstype => 'float4' },
+ aggcombinefn => 'float4pl', aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +84,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialconverterfn => 'numeric_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..2c63102ff31 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* function to convert aggregate result to bytea (0 if none) */
+ regproc aggpartialconverterfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -91,6 +94,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -161,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -175,6 +182,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index d068d6532ec..1cf23b15df0 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -4798,6 +4798,12 @@
{ oid => '2786', descr => 'aggregate serial function',
proname => 'int8_avg_serialize', prorettype => 'bytea',
proargtypes => 'internal', prosrc => 'int8_avg_serialize' },
+{ oid => '9630', descr => 'partial aggregate converter function',
+ proname => 'int8_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'int8_sum_to_internal_serialize' },
+{ oid => '9631', descr => 'partial aggregate converter function',
+ proname => 'numeric_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'numeric_sum_to_internal_serialize' },
{ oid => '2787', descr => 'aggregate deserial function',
proname => 'int8_avg_deserialize', prorettype => 'internal',
proargtypes => 'bytea internal', prosrc => 'int8_avg_deserialize' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be3..864151bbca3 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {aggpartialconverterfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2021-11-15 10:16 ` Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Daniel Gustafsson @ 2021-11-15 10:16 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Ilya Gladyshev <[email protected]>; pgsql-hackers
> On 3 Nov 2021, at 15:50, Alexander Pyhalov <[email protected]> wrote:
>
> Daniel Gustafsson писал 2021-11-03 16:45:
>>> On 2 Nov 2021, at 10:12, Alexander Pyhalov <[email protected]> wrote:
>>> Updated and rebased patch.
>> + state = (Int128AggState *) palloc0(sizeof(Int128AggState));
>> + state->calcSumX2 = false;
>> +
>> + if (!PG_ARGISNULL(0))
>> + {
>> +#ifdef HAVE_INT128
>> + do_int128_accum(state, (int128) PG_GETARG_INT64(0));
>> +#else
>> + do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
>> +#endif
>> This fails on non-INT128 platforms as state cannot be cast to Int128AggState
>> outside of HAVE_INT128; it's not defined there. This needs to be a
>> PolyNumAggState no?
>
> Hi.
> Thank you for noticing this. It's indeed fails with pgac_cv__128bit_int=no.
> Updated patch.
The updated patch also fails to apply now, but on the catversion.h bump. To
avoid having to rebase for that I recommend to skip that part in the patch and
just mention the need in the thread, any committer picking this up for commit
will know to bump the catversion so there is no use in risking unneccesary
conflicts.
--
Daniel Gustafsson https://vmware.com/
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
@ 2021-11-15 13:01 ` Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2021-11-15 13:01 UTC (permalink / raw)
To: Daniel Gustafsson <[email protected]>; +Cc: Ilya Gladyshev <[email protected]>; pgsql-hackers
Daniel Gustafsson писал 2021-11-15 13:16:
>> On 3 Nov 2021, at 15:50, Alexander Pyhalov <[email protected]>
>> wrote:
>>
>> Daniel Gustafsson писал 2021-11-03 16:45:
>>>> On 2 Nov 2021, at 10:12, Alexander Pyhalov
>>>> <[email protected]> wrote:
>>>> Updated and rebased patch.
>>> + state = (Int128AggState *) palloc0(sizeof(Int128AggState));
>>> + state->calcSumX2 = false;
>>> +
>>> + if (!PG_ARGISNULL(0))
>>> + {
>>> +#ifdef HAVE_INT128
>>> + do_int128_accum(state, (int128) PG_GETARG_INT64(0));
>>> +#else
>>> + do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
>>> +#endif
>>> This fails on non-INT128 platforms as state cannot be cast to
>>> Int128AggState
>>> outside of HAVE_INT128; it's not defined there. This needs to be a
>>> PolyNumAggState no?
>>
>> Hi.
>> Thank you for noticing this. It's indeed fails with
>> pgac_cv__128bit_int=no.
>> Updated patch.
>
> The updated patch also fails to apply now, but on the catversion.h
> bump. To
> avoid having to rebase for that I recommend to skip that part in the
> patch and
> just mention the need in the thread, any committer picking this up for
> commit
> will know to bump the catversion so there is no use in risking
> unneccesary
> conflicts.
I've updated patch - removed catversion dump.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v07.patch (57.7K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v07.patch)
download | inline diff:
From 2af16e66276938b861cf7a8db2fef967f54b800f Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 57 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 185 ++++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 196 +++++++++++++++++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 27 ++-
src/backend/catalog/pg_aggregate.c | 28 ++-
src/backend/commands/aggregatecmds.c | 23 +-
src/backend/utils/adt/numeric.c | 96 +++++++++
src/bin/pg_dump/pg_dump.c | 21 +-
src/include/catalog/pg_aggregate.dat | 106 +++++-----
src/include/catalog/pg_aggregate.h | 10 +-
src/include/catalog/pg_proc.dat | 6 +
src/test/regress/expected/oidjoins.out | 1 +
12 files changed, 673 insertions(+), 83 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index b27689d0864..a515c5662bb 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -197,6 +197,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -832,8 +833,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3349,7 +3352,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3819,3 +3822,51 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (!aggform->aggpartialpushdownsafe)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /*
+ * If an aggregate requires serialization/deserialization, partial
+ * converter should be defined
+ */
+ if (agg->aggtranstype == INTERNALOID && aggform->aggpartialconverterfn == InvalidOid)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ /* In this case we currently don't use converter */
+ if (agg->aggtranstype != INTERNALOID && get_func_rettype(agg->aggfnoid) != agg->aggtranstype)
+ {
+ ReleaseSysCache(aggtup);
+ return false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return true;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 3cee0a8c12b..6ffaf158c06 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9326,13 +9326,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9391,8 +9391,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9403,21 +9403,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9453,6 +9453,173 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*), sum(pagg_tab.d)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ max | count | sum
+-----+-------+-------
+ 29 | 3000 | 58500
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..bad03c49e49 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_class.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/vacuum.h"
@@ -48,6 +50,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -80,7 +83,12 @@ enum FdwScanPrivateIndex
* String describing join i.e. names of relations being joined and types
* of join, added when the scan is join
*/
- FdwScanPrivateRelations
+ FdwScanPrivateRelations,
+
+ /*
+ * List of functions to convert partial aggregate result
+ */
+ FdwScanPrivateConvertors
};
/*
@@ -143,6 +151,7 @@ typedef struct PgFdwScanState
/* extracted fdw_private data */
char *query; /* text of SELECT command */
List *retrieved_attrs; /* list of retrieved attribute numbers */
+ List *conv_list; /* list of converters */
/* for remote query execution */
PGconn *conn; /* connection for the scan */
@@ -474,6 +483,7 @@ static void store_returning_result(PgFdwModifyState *fmstate,
TupleTableSlot *slot, PGresult *res);
static void finish_foreign_modify(PgFdwModifyState *fmstate);
static void deallocate_query(PgFdwModifyState *fmstate);
+static List *build_conv_list(RelOptInfo *foreignrel);
static List *build_remote_returning(Index rtindex, Relation rel,
List *returningList);
static void rebuild_fdw_scan_tlist(ForeignScan *fscan, List *tlist);
@@ -510,6 +520,7 @@ static HeapTuple make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context);
static void conversion_error_callback(void *arg);
@@ -517,7 +528,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +537,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -541,7 +553,6 @@ static void merge_fdw_options(PgFdwRelationInfo *fpinfo,
const PgFdwRelationInfo *fpinfo_i);
static int get_batch_size_option(Relation rel);
-
/*
* Foreign-data wrapper handler function: return a struct with pointers
* to my callback routines.
@@ -1233,6 +1244,7 @@ postgresGetForeignPlan(PlannerInfo *root,
List *local_exprs = NIL;
List *params_list = NIL;
List *fdw_scan_tlist = NIL;
+ List *fdw_conv_list = NIL;
List *fdw_recheck_quals = NIL;
List *retrieved_attrs;
StringInfoData sql;
@@ -1336,6 +1348,9 @@ postgresGetForeignPlan(PlannerInfo *root,
/* Build the list of columns to be fetched from the foreign server. */
fdw_scan_tlist = build_tlist_to_deparse(foreignrel);
+ /* Build the list of converters for partial aggregates. */
+ fdw_conv_list = build_conv_list(foreignrel);
+
/*
* Ensure that the outer plan produces a tuple whose descriptor
* matches our scan tuple slot. Also, remove the local conditions
@@ -1415,6 +1430,8 @@ postgresGetForeignPlan(PlannerInfo *root,
if (IS_JOIN_REL(foreignrel) || IS_UPPER_REL(foreignrel))
fdw_private = lappend(fdw_private,
makeString(fpinfo->relation_name));
+ if (IS_UPPER_REL(foreignrel))
+ fdw_private = lappend(fdw_private, fdw_conv_list);
/*
* Create the ForeignScan node for the given relation.
@@ -1433,6 +1450,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expected to return BYTEA, but real input type is likely different.
+ */
+static AttInMetadata *
+get_rcvd_attinmeta(TupleDesc tupdesc, List *conv_list)
+{
+ TupleDesc rcvd_tupdesc;
+
+ Assert(conv_list != NIL);
+
+ rcvd_tupdesc = CreateTupleDescCopy(tupdesc);
+ for (int i = 0; i < rcvd_tupdesc->natts; i++)
+ {
+ Oid converter = InvalidOid;
+ Form_pg_attribute att = TupleDescAttr(rcvd_tupdesc, i);
+
+ converter = list_nth_oid(conv_list, i);
+ if (converter != InvalidOid)
+ {
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(converter));
+
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", converter);
+
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ if (procform->pronargs != 1)
+ elog(ERROR, "converter %s is expected to have one argument", NameStr(procform->proname));
+
+ att->atttypid = procform->proargtypes.values[0];
+
+ ReleaseSysCache(proctup);
+ }
+ }
+
+ return TupleDescGetAttInMetadata(rcvd_tupdesc);
+}
+
/*
* Construct a tuple descriptor for the scan tuples handled by a foreign join.
*/
@@ -1545,6 +1604,12 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
FdwScanPrivateSelectSql));
fsstate->retrieved_attrs = (List *) list_nth(fsplan->fdw_private,
FdwScanPrivateRetrievedAttrs);
+
+ if (list_length(fsplan->fdw_private) > FdwScanPrivateConvertors)
+ fsstate->conv_list = (List *) list_nth(fsplan->fdw_private,
+ FdwScanPrivateConvertors);
+
+
fsstate->fetch_size = intVal(list_nth(fsplan->fdw_private,
FdwScanPrivateFetchSize));
@@ -1571,7 +1636,10 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
fsstate->tupdesc = get_tupdesc_for_join_scan_tuples(node);
}
- fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
+ if (fsstate->conv_list != NIL)
+ fsstate->attinmeta = get_rcvd_attinmeta(fsstate->tupdesc, fsstate->conv_list);
+ else
+ fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
/*
* Prepare for processing of parameters used in remote query, if any.
@@ -3822,6 +3890,7 @@ fetch_more_data(ForeignScanState *node)
fsstate->rel,
fsstate->attinmeta,
fsstate->retrieved_attrs,
+ fsstate->conv_list,
node,
fsstate->temp_cxt);
}
@@ -4309,6 +4378,7 @@ store_returning_result(PgFdwModifyState *fmstate,
fmstate->rel,
fmstate->attinmeta,
fmstate->retrieved_attrs,
+ NIL,
NULL,
fmstate->temp_cxt);
@@ -4603,6 +4673,7 @@ get_returning_data(ForeignScanState *node)
dmstate->rel,
dmstate->attinmeta,
dmstate->retrieved_attrs,
+ NIL,
node,
dmstate->temp_cxt);
ExecStoreHeapTuple(newtup, slot, false);
@@ -5183,6 +5254,7 @@ analyze_row_processor(PGresult *res, int row, PgFdwAnalyzeState *astate)
astate->rel,
astate->attinmeta,
astate->retrieved_attrs,
+ NIL,
NULL,
astate->temp_cxt);
@@ -6083,7 +6155,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6169,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6413,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6428,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6457,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6474,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6496,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
@@ -7108,6 +7192,30 @@ complete_pending_request(AsyncRequest *areq)
TupIsNull(areq->result) ? 0.0 : 1.0);
}
+/*
+ * Interface to fmgr to call converters
+ */
+static Datum
+call_converter(Oid converter, Oid collation, Datum value, bool isnull, bool *res_isnull)
+{
+ LOCAL_FCINFO(fcinfo, 1);
+ FmgrInfo flinfo;
+ Datum result;
+
+ fmgr_info(converter, &flinfo);
+
+ InitFunctionCallInfoData(*fcinfo, &flinfo, 1, collation, NULL, NULL);
+
+ fcinfo->args[0].value = value;
+ fcinfo->args[0].isnull = isnull;
+
+ result = FunctionCallInvoke(fcinfo);
+
+ if (res_isnull)
+ *res_isnull = fcinfo->isnull;
+ return result;
+}
+
/*
* Create a tuple from the specified row of the PGresult.
*
@@ -7127,6 +7235,7 @@ make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context)
{
@@ -7209,6 +7318,20 @@ make_tuple_from_result_row(PGresult *res,
valstr,
attinmeta->attioparams[i - 1],
attinmeta->atttypmods[i - 1]);
+ if (conv_list != NIL)
+ {
+ Oid converter = list_nth_oid(conv_list, i - 1);
+
+ if (converter != InvalidOid)
+ {
+ Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+ bool res_isnull;
+
+ values[i - 1] = call_converter(converter, att->attcollation, values[i - 1], nulls[i - 1], &res_isnull);
+
+ nulls[i - 1] = res_isnull;
+ }
+ }
}
else if (i == SelfItemPointerAttributeNumber)
{
@@ -7472,3 +7595,54 @@ get_batch_size_option(Relation rel)
return batch_size;
}
+
+/*
+ * For UPPER_REL build a list of converters, corresponding to tlist entries.
+ */
+static List *
+build_conv_list(RelOptInfo *foreignrel)
+{
+ List *conv_list = NIL;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) foreignrel->fdw_private;
+ ListCell *lc;
+
+ if (IS_UPPER_REL(foreignrel))
+ {
+ /* For UPPER_REL tlist matches grouped_tlist */
+ foreach(lc, fpinfo->grouped_tlist)
+ {
+ TargetEntry *tlentry = (TargetEntry *) lfirst(lc);
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
+ {
+ Aggref *agg = (Aggref *) tlentry->expr;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && agg->aggtranstype == INTERNALOID)
+ {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ converter_oid = aggform->aggpartialconverterfn;
+ Assert(converter_oid != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+ }
+
+ /*
+ * We append InvalidOid to conv_list to preserve one-to-one
+ * mapping between tlist and conv_list members.
+ */
+ conv_list = lappend_oid(conv_list, converter_oid);
+ }
+ }
+
+ return conv_list;
+}
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index e40112e41d3..36fe8a69bcd 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2747,15 +2747,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2789,6 +2789,25 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..639ea4cf9a6 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -60,6 +60,7 @@ AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -74,7 +75,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -88,6 +90,7 @@ AggregateCreate(const char *aggName,
Oid combinefn = InvalidOid; /* can be omitted */
Oid serialfn = InvalidOid; /* can be omitted */
Oid deserialfn = InvalidOid; /* can be omitted */
+ Oid partialconverterfn = InvalidOid; /* can be omitted */
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
@@ -569,6 +572,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial converter, if present.
+ */
+ if (aggpartialconverterfnName)
+ {
+ fnArgs[0] = finaltype;
+
+ partialconverterfn = lookup_agg_function(aggpartialconverterfnName, 1,
+ fnArgs, InvalidOid,
+ &rettype);
+
+ if (rettype != BYTEAOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial serialization function %s is not %s",
+ NameListToString(aggserialfnName),
+ format_type_be(BYTEAOID))));
+
+ }
+
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -664,6 +688,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggcombinefn - 1] = ObjectIdGetDatum(combinefn);
values[Anum_pg_aggregate_aggserialfn - 1] = ObjectIdGetDatum(serialfn);
values[Anum_pg_aggregate_aggdeserialfn - 1] = ObjectIdGetDatum(deserialfn);
+ values[Anum_pg_aggregate_aggpartialconverterfn - 1] = ObjectIdGetDatum(partialconverterfn);
values[Anum_pg_aggregate_aggmtransfn - 1] = ObjectIdGetDatum(mtransfn);
values[Anum_pg_aggregate_aggminvtransfn - 1] = ObjectIdGetDatum(minvtransfn);
values[Anum_pg_aggregate_aggmfinalfn - 1] = ObjectIdGetDatum(mfinalfn);
@@ -676,6 +701,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..3b7c1597315 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -69,11 +69,13 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialconverterfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -142,6 +144,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialconverterfunc") == 0)
+ partialconverterfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -189,6 +193,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -372,6 +378,18 @@ DefineAggregate(ParseState *pstate,
errmsg("must specify both or neither of serialization and deserialization functions")));
}
+ if (partialconverterfuncName)
+ {
+ /*
+ * Converter is only needed/allowed for transtype INTERNAL.
+ */
+ if (transTypeId != INTERNALOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
+ errmsg("partial converters may be specified only when the aggregate transition data type is %s",
+ format_type_be(INTERNALOID))));
+ }
+
/*
* If a moving-aggregate transtype is specified, look that up. Same
* restrictions as for transtype.
@@ -457,6 +475,8 @@ DefineAggregate(ParseState *pstate,
combinefuncName, /* combine function name */
serialfuncName, /* serial function name */
deserialfuncName, /* deserial function name */
+ partialconverterfuncName, /* partial converter function
+ * name */
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
@@ -471,7 +491,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 1de744855f3..ac900867e87 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -5770,6 +5770,52 @@ int8_avg_deserialize(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
+/*
+ * int8_sum_to_internal_serialize
+ * Convert int8 argument to serialized internal representation
+ */
+Datum
+int8_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ PolyNumAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = (PolyNumAggState *) palloc0(sizeof(PolyNumAggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
+ }
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+#ifdef HAVE_INT128
+ int128_to_numericvar(state->sumX, &tmp_var);
+#else
+ accum_sum_final(&state->sumX, &tmp_var);
+#endif
+ numericvar_serialize(&buf, &tmp_var);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Inverse transition functions to go with the above.
*/
@@ -5995,6 +6041,56 @@ numeric_sum(PG_FUNCTION_ARGS)
PG_RETURN_NUMERIC(result);
}
+/*
+ * numeric_sum_to_internal_serialize
+ * Convert numeric argument to serialized internal representation
+ */
+Datum
+numeric_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ NumericAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = makeNumericAggStateCurrentContext(false);
+
+ if (!PG_ARGISNULL(0))
+ do_numeric_accum(state, PG_GETARG_NUMERIC(0));
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+ accum_sum_final(&state->sumX, &tmp_var);
+ numericvar_serialize(&buf, &tmp_var);
+
+ /* maxScale */
+ pq_sendint32(&buf, state->maxScale);
+
+ /* maxScaleCount */
+ pq_sendint64(&buf, state->maxScaleCount);
+
+ /* NaNcount */
+ pq_sendint64(&buf, state->NaNcount);
+
+ /* pInfcount */
+ pq_sendint64(&buf, state->pInfcount);
+
+ /* nInfcount */
+ pq_sendint64(&buf, state->nInfcount);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Workhorse routine for the standard deviance and variance
* aggregates. 'state' is aggregate's transition state.
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index 7e98371d253..877fde809f1 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -14076,11 +14076,13 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggcombinefn;
const char *aggserialfn;
const char *aggdeserialfn;
+ const char *aggpartialconverterfn;
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
bool aggfinalextra;
bool aggmfinalextra;
+ bool aggpartialpushdownsafe;
char aggfinalmodify;
char aggmfinalmodify;
const char *aggsortop;
@@ -14165,11 +14167,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 150000)
+ appendPQExpBufferStr(query,
+ "aggpartialconverterfn,\n"
+ "aggpartialpushdownsafe\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS aggpartialconverterfn,\n"
+ "false as aggpartialpushdownsafe\n");
appendPQExpBuffer(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -14187,6 +14197,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ aggpartialconverterfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialconverterfn"));
+ aggpartialpushdownsafe = (PQgetvalue(res, 0, PQfnumber(res, "aggpartialpushdownsafe"))[0] == 't');
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -14281,6 +14293,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(aggpartialconverterfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALCONVERTERFUNC = %s", aggpartialconverterfn);
+ if (aggpartialpushdownsafe)
+ appendPQExpBufferStr(details, ",\n PARTIAL_PUSHDOWN_SAFE");
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..61c4d812b34 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -56,26 +56,27 @@
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ aggpartialconverterfn => 'int8_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
- aggcombinefn => 'float4pl', aggtranstype => 'float4' },
+ aggcombinefn => 'float4pl', aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +84,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialconverterfn => 'numeric_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..2c63102ff31 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* function to convert aggregate result to bytea (0 if none) */
+ regproc aggpartialconverterfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -91,6 +94,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -161,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -175,6 +182,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index d068d6532ec..1cf23b15df0 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -4798,6 +4798,12 @@
{ oid => '2786', descr => 'aggregate serial function',
proname => 'int8_avg_serialize', prorettype => 'bytea',
proargtypes => 'internal', prosrc => 'int8_avg_serialize' },
+{ oid => '9630', descr => 'partial aggregate converter function',
+ proname => 'int8_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'int8_sum_to_internal_serialize' },
+{ oid => '9631', descr => 'partial aggregate converter function',
+ proname => 'numeric_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'numeric_sum_to_internal_serialize' },
{ oid => '2787', descr => 'aggregate deserial function',
proname => 'int8_avg_deserialize', prorettype => 'internal',
proargtypes => 'bytea internal', prosrc => 'int8_avg_deserialize' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be3..864151bbca3 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {aggpartialconverterfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2022-08-01 05:55 ` [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2022-08-01 05:55 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; +Cc: Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; pgsql-hackers; [email protected] <[email protected]>
Hi Mr.Vondra, Mr.Pyhalov.
I'm interesied in Mr.Pyhalov's patch due to the following background.
--Background
I develop postgresql's extension such as fdw in my work.
I'm interested in using postgresql for OLAP.
I think the function of a previous patch "Push aggregation down to base relations and joins"[1] is desiable. I rebased the previous patch and register the rebased patch on the next commitfest[2].
And I think it would be more useful if the previous patch works on a foreign table of postgres_fdw.
I realized the function of partial aggregation pushdown is necessary to make the previous patch work on a foreign table of postgres_fdw.
--
So I reviewed Mr.Pyhalov's patch and discussions on this thread.
I made a draft of approach to respond to Mr.Vondra's comments.
Would you check whether my draft is right or not?
--My draft
> 1) It's not clear to me how could this get extended to aggregates with
> more complex aggregate states, to support e.g. avg() and similar
> fairly common aggregates.
We add a special aggregate function every aggregate function (hereafter we call this src) which supports partial aggregation.
The followings are differences between the src and the special aggregate function.
difference1) result type
The result type is same with the src's transtype if the src's transtype is not internal.
Otherwise the result type is bytea.
difference2) final func
The final func does not exist if the src's transtype is not internal.
Otherwize the final func returns serialized value.
For example, let me call the special aggregate function of avg(float8) avg_p(float8).
The result value of avg_p is a float8 array which consists of count and summation.
avg_p does not have finalfunc.
We pushdown the special aggregate function instead of a src.
For example, we issue "select avg_p(c) from t" instead of "select avg(c) from t"
in the above example.
We add a new column partialaggfn to pg_aggregate to get the oid of the special aggregate function from the the src's oid.
This column is the oid of the special aggregate function which corresponds to the src.
If an aggregate function does not have any special aggregate function, then we does not pushdown any partial aggregation of the aggregate function.
> 2) I'm not sure relying on aggpartialpushdownsafe without any version
> checks etc. is sufficient. I mean, how would we know the remote node
> has the same idea of representing the aggregate state. I wonder how
> this aligns with assumptions we do e.g. for functions etc.
We add compatible server versions infomation to pg_aggregate and the set of options of postgres_fdw's foreign server.
We check compatibility of an aggregate function using this infomation.
An additional column of pg_aggregate is compatibleversonrange.
This column is a range of postgresql server versions which has compatible aggregate function.
An additional options of postgres_fdw's foreign server are serverversion and bwcompatibleverson.
serverversion is remote postgresql server version.
bwcompatibleverson is the maximum version in which any aggregate function is compatible with local noed's one.
Our version check passes if and only if at least one of the following conditions is true.
condition1) the option value of serverversion is in compatibleversonrange.
condition2) the local postgresql server version is between bwcompatibleverson and the option value of serverversion.
We can get the local postgresql server version from PG_VERSION_NUM macro.
We use condition1 if the local postgresql server version is not more than the remote one.
and use condition2 if the local postgresql server version is greater than the remote one.
--
Sincerely yours,
Yuuki Fujii
[1] https://commitfest.postgresql.org/32/1247/
[2] https://commitfest.postgresql.org/39/3764/
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-22 01:01 ` [email protected] <[email protected]>
2022-11-22 05:00 ` Re: Partial aggregates pushdown Ted Yu <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 2 replies; 42+ messages in thread
From: [email protected] @ 2022-11-22 01:01 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; +Cc: Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Vondra, Mr.Pyhalov, Everyone.
I discussed with Mr.Pyhalov about the above draft by directly sending mail to
him(outside of pgsql-hackers). Mr.Pyhalov allowed me to update his patch
along with the above draft. So I update Mr.Pyhalov's patch v10.
I wrote my patch for discussion.
My patch passes regression tests which contains additional basic postgres_fdw tests
for my patch's feature. But my patch doesn't contain sufficient documents and tests.
If reviewers accept my approach, I will add documents and tests to my patch.
The following is a my patch's readme.
# I simplified the above draft.
--readme of my patch
1. interface
1) pg_aggregate
There are the following additional columns.
a) partialaggfn
data type : regproc.
default value: zero(means invalid).
description : This field refers to the special aggregate function(then we call
this partialaggfunc)
corresponding to aggregation function(then we call src) which has aggfnoid.
partialaggfunc is used for partial aggregation pushdown by postgres_fdw.
The followings are differences between the src and the special aggregate function.
difference1) result type
The result type is same as the src's transtype if the src's transtype
is not internal.
Otherwise the result type is bytea.
difference2) final func
The final func does not exist if the src's transtype is not internal.
Otherwize the final func returns serialized value.
For example, there is a partialaggfunc avg_p_int4 which corresponds to avg(int4)
whose aggtranstype is _int4.
The result value of avg_p_int4 is a float8 array which consists of count and
summation. avg_p_int4 does not have finalfunc.
For another example, there is a partialaggfunc avg_p_int8 which corresponds to
avg(int8) whose aggtranstype is internal.
The result value of avg_p_int8 is a bytea serialized array which consists of count
and summation. avg_p_int8 has finalfunc int8_avg_serialize which is serialize function
of avg(int8). This field is zero if there is no partialaggfunc.
b) partialagg_minversion
data type : int4.
default value: zero(means current version).
description : This field is the minimum PostgreSQL server version which has
partialaggfunc. This field is used for checking compatibility of partialaggfunc.
The above fields are valid in tuples for builtin avg, sum, min, max, count.
There are additional records which correspond to partialaggfunc for avg, sum, min, max,
count.
2) pg_proc
There are additional records which correspond to partialaggfunc for avg, sum, min, max,
count.
3) postgres_fdw
postgres_fdw has an additional foreign server option server_version. server_version is
integer value which means remote server version number. Default value of server_version
is zero. server_version is used for checking compatibility of partialaggfunc.
2. feature
postgres_fdw can pushdown partial aggregation of avg, sum, min, max, count.
Partial aggregation pushdown is fine when the following two conditions are both true.
condition1) partialaggfn is valid.
condition2) server_version is not less than partialagg_minversion
postgres_fdw executes pushdown the patialaggfunc instead of a src.
For example, we issue "select avg_p_int4(c) from t" instead of "select avg(c) from t"
in the above example.
postgres_fdw can pushdown every aggregate function which supports partial aggregation
if you add a partialaggfunc corresponding to the aggregate function by create aggregate
command.
3. difference between my patch and Mr.Pyhalov's v10 patch.
1) In my patch postgres_fdw can pushdown partial aggregation of avg
2) In my patch postgres_fdw can pushdown every aggregate function which supports partial
aggregation if you add a partialaggfunc corresponding to the aggregate function.
4. sample commands in psql
\c postgres
drop database tmp;
create database tmp;
\c tmp
create extension postgres_fdw;
create server server_01 foreign data wrapper postgres_fdw options(host 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
create user mapping for postgres server server_01 options(user 'postgres', password 'postgres');
create server server_02 foreign data wrapper postgres_fdw options(host 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
create user mapping for postgres server server_02 options(user 'postgres', password 'postgres');
create table t(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval) partition by list (type);
create table t1(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval);
create table t2(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval);
truncate table t1;
truncate table t2;
insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 1, 1.1, 1, cast('1 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 2, 2.1, 1, cast('2 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 1, 1.1, 2, cast('1 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 2, 2.1, 2, cast('2 seconds' as interval) from generate_series(1, 100000, 1) t;
create foreign table f_t1 partition of t for values in (1) server server_01 options(table_name 't1');
create foreign table f_t2 partition of t for values in (2) server server_02 options(table_name 't2');
set enable_partitionwise_aggregate = on;
explain (verbose, costs off) select avg(total::int4), avg(total::int8) from t;
select avg(total::int4), avg(total::int8) from t;
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v11.patch (107.1K, ../../OS3PR01MB66607C2ABBDAEE150F32E24C950D9@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v11.patch)
download | inline diff:
From dcdf4d107dc1d062f54e91abfac7099956b17988 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Sat, 19 Nov 2022 18:00:01 +0900
Subject: [PATCH] Partial aggregates push down v11
---
contrib/postgres_fdw/deparse.c | 91 ++++-
.../postgres_fdw/expected/postgres_fdw.out | 246 +++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 25 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 54 ++-
src/backend/catalog/pg_aggregate.c | 50 ++-
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 367 +++++++++++++++---
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 207 ++++++++++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/expected/opr_sanity.out | 120 +++++-
14 files changed, 1117 insertions(+), 115 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..3341615737 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,33 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+ if (aggform->partialaggfn) {
+ appendFunctionName((Oid)(aggform->partialaggfn), context);
+ } else {
+ elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3981,60 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg has compatibility
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ bool compatible = false;
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion == PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = PG_VERSION_NUM;
+ }
+ else {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ /* the option value of serverversion is in compatibleversonrange */
+ if ((fpinfo->server_version >= partialagg_minversion)) {
+ compatible = true;
+ }
+ return compatible;
+}
+/*
+ * Check that partial aggregate agg is fine to push down
+ *
+ * Check that it's AGGKIND_NORMAL aggregate without
+ * distinct or order by clauses, which has valid partialaggfn
+ * and partialagg_minversion <= server_version
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn)
+ {
+ partial_agg_ok = false;
+ } else if(!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..cf0f270946 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS name AS $$
+ SELECT (setting::int4 + version_offset)::name FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name name, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num name;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,213 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | avg | avg
+----+-----+-----+-------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | avg | avg
+-----+-----+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(name, text, int4);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..c8efdfefb7 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -527,7 +527,7 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra, bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -6159,7 +6162,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6176,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6416,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6423,7 +6431,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6448,7 +6460,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra, bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6477,6 +6489,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
fpinfo->table = ifpinfo->table;
fpinfo->server = ifpinfo->server;
fpinfo->user = ifpinfo->user;
+ fpinfo->server_version = ifpinfo->server_version;
merge_fdw_options(fpinfo, ifpinfo, NULL);
/*
@@ -6485,7 +6498,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..5bb35d9743 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for check compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..9176337420 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS name AS $$
+ SELECT (setting::int4 + version_offset)::name FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name name, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num name;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,34 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(name, text, int4);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..9f8a326cec 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -36,7 +36,8 @@
static Oid lookup_agg_function(List *fnName, int nargs, Oid *input_types,
Oid variadicArgType,
- Oid *rettype);
+ Oid *rettype,
+ bool only_normal);
/*
@@ -63,6 +64,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +74,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +94,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -228,7 +232,7 @@ AggregateCreate(const char *aggName,
}
transfn = lookup_agg_function(aggtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/*
* Return type of transfn (possibly after refinement by
@@ -281,7 +285,7 @@ AggregateCreate(const char *aggName,
mtransfn = lookup_agg_function(aggmtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/* As above, return type must exactly match declared mtranstype. */
if (rettype != aggmTransType)
@@ -326,7 +330,7 @@ AggregateCreate(const char *aggName,
minvtransfn = lookup_agg_function(aggminvtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/* As above, return type must exactly match declared mtranstype. */
if (rettype != aggmTransType)
@@ -382,7 +386,7 @@ AggregateCreate(const char *aggName,
finalfn = lookup_agg_function(aggfinalfnName, nargs_finalfn,
fnArgs, ffnVariadicArgType,
- &finaltype);
+ &finaltype, true);
/*
* When finalfnExtraArgs is specified, the finalfn will certainly be
@@ -418,7 +422,7 @@ AggregateCreate(const char *aggName,
combinefn = lookup_agg_function(aggcombinefnName, 2,
fnArgs, InvalidOid,
- &combineType);
+ &combineType, true);
/* Ensure the return type matches the aggregate's trans type */
if (combineType != aggTransType)
@@ -450,7 +454,7 @@ AggregateCreate(const char *aggName,
serialfn = lookup_agg_function(aggserialfnName, 1,
fnArgs, InvalidOid,
- &rettype);
+ &rettype, true);
if (rettype != BYTEAOID)
ereport(ERROR,
@@ -471,7 +475,7 @@ AggregateCreate(const char *aggName,
deserialfn = lookup_agg_function(aggdeserialfnName, 2,
fnArgs, InvalidOid,
- &rettype);
+ &rettype, true);
if (rettype != INTERNALOID)
ereport(ERROR,
@@ -545,7 +549,7 @@ AggregateCreate(const char *aggName,
mfinalfn = lookup_agg_function(aggmfinalfnName, nargs_finalfn,
fnArgs, ffnVariadicArgType,
- &rettype);
+ &rettype, true);
/* As above, check strictness if mfinalfnExtraArgs is given */
if (mfinalfnExtraArgs && func_strict(mfinalfn))
@@ -569,6 +573,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = lookup_agg_function(partialaggfnName, numArgs,
+ aggArgTypes, variadicArgType, &rettype, false);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial aggregation function %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +709,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
@@ -827,7 +854,8 @@ lookup_agg_function(List *fnName,
int nargs,
Oid *input_types,
Oid variadicArgType,
- Oid *rettype)
+ Oid *rettype,
+ bool only_normal)
{
Oid fnOid;
bool retset;
@@ -852,7 +880,7 @@ lookup_agg_function(List *fnName,
&true_oid_array, NULL);
/* only valid case is a normal function not returning a set */
- if (fdresult != FUNCDETAIL_NORMAL || !OidIsValid(fnOid))
+ if ((fdresult != FUNCDETAIL_NORMAL || !OidIsValid(fnOid)) && only_normal)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("function %s does not exist",
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..fb1cceeb62 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,46 +58,87 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_int4', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}',
+ partialaggfn => 'sum_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_int2', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
-{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
+ aggmtranstype => '_int8', aggminitval => '{0,0}',
+ partialaggfn => 'sum_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_float4', aggtransfn => 'float4pl',
aggcombinefn => 'float4pl', aggtranstype => 'float4' },
-{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
+{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
+ aggcombinefn => 'float4pl', aggtranstype => 'float4',
+ partialaggfn => 'sum_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_float8', aggtransfn => 'float8pl',
aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
+ aggcombinefn => 'float8pl', aggtranstype => 'float8',
+ partialaggfn => 'sum_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_money', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
+ aggtranstype => 'money' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money',
+ partialaggfn => 'sum_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_interval', aggtransfn => 'interval_pl',
+ aggcombinefn => 'interval_pl', aggtranstype => 'interval' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval',
+ partialaggfn => 'sum_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,152 +146,336 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
-{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
+{ aggfnoid => 'max_p_int8', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
aggtranstype => 'int8' },
-{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
+{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
+ aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
+ aggtranstype => 'int8',
+ partialaggfn => 'max_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_int4', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
aggtranstype => 'int4' },
-{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
+{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
+ aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
+ aggtranstype => 'int4',
+ partialaggfn => 'max_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_int2', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
aggtranstype => 'int2' },
-{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
+{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
+ aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
+ aggtranstype => 'int2',
+ partialaggfn => 'max_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_oid', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
aggtranstype => 'oid' },
-{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
+{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
+ aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
+ aggtranstype => 'oid',
+ partialaggfn => 'max_p_oid', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_float4', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
aggtranstype => 'float4' },
-{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
+{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
+ aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
+ aggtranstype => 'float4',
+ partialaggfn => 'max_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_floa8', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
aggtranstype => 'float8' },
-{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
+{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
+ aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
+ aggtranstype => 'float8',
+ partialaggfn => 'max_p_floa8', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_date', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
aggtranstype => 'date' },
-{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
+{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
+ aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
+ aggtranstype => 'date',
+ partialaggfn => 'max_p_date', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_time', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
aggtranstype => 'time' },
-{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
+{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
+ aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
+ aggtranstype => 'time',
+ partialaggfn => 'max_p_time', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timetz', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
aggtranstype => 'timetz' },
-{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
+{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
+ aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
+ aggtranstype => 'timetz',
+ partialaggfn => 'max_p_timetz', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_money', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
aggtranstype => 'money' },
-{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
+{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
+ aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
+ aggtranstype => 'money',
+ partialaggfn => 'max_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timestamp', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
aggtranstype => 'timestamp' },
-{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
+{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
+ aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
+ aggtranstype => 'timestamp',
+ partialaggfn => 'max_p_timestamp', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timestamptz', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
-{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
+{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
+ aggcombinefn => 'timestamptz_larger',
+ aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ partialaggfn => 'max_p_timestamptz', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_interval', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
aggtranstype => 'interval' },
-{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
+{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
+ aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
+ aggtranstype => 'interval',
+ partialaggfn => 'max_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_text', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
aggtranstype => 'text' },
-{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
+{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
+ aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
+ aggtranstype => 'text',
+ partialaggfn => 'max_p_text', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_numeric', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
aggtranstype => 'numeric' },
-{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
+{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
+ aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
+ aggtranstype => 'numeric',
+ partialaggfn => 'max_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_anyarray', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
aggtranstype => 'anyarray' },
-{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
+{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
+ aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
+ aggtranstype => 'anyarray',
+ partialaggfn => 'max_p_anyarray', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_bpchar', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
aggtranstype => 'bpchar' },
-{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
+{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
+ aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
+ aggtranstype => 'bpchar',
+ partialaggfn => 'max_p_bpchar', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_tid', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
aggtranstype => 'tid' },
-{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
+{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
+ aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
+ aggtranstype => 'tid',
+ partialaggfn => 'max_p_tid', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_anyenum', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
aggtranstype => 'anyenum' },
-{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
+{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
+ aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
+ aggtranstype => 'anyenum',
+ partialaggfn => 'max_p_anyenum', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_inet', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
aggtranstype => 'inet' },
-{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
+{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
+ aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
+ aggtranstype => 'inet',
+ partialaggfn => 'max_p_inet', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_pg_lsn', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
aggtranstype => 'pg_lsn' },
-{ aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger',
+{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
+ aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
+ aggtranstype => 'pg_lsn',
+ partialaggfn => 'max_p_pg_lsn', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_xid8', aggtransfn => 'xid8_larger',
aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)',
aggtranstype => 'xid8' },
+{ aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger',
+ aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)',
+ aggtranstype => 'xid8',
+ partialaggfn => 'max_p_xid8', partialagg_minversion => '160000' },
# min
-{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
+{ aggfnoid => 'min_p_int8', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
aggtranstype => 'int8' },
-{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
+{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
+ aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
+ aggtranstype => 'int8',
+ partialaggfn => 'min_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_int4', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
aggtranstype => 'int4' },
-{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
+{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
+ aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
+ aggtranstype => 'int4',
+ partialaggfn => 'min_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_int2', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
aggtranstype => 'int2' },
-{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
+{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
+ aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
+ aggtranstype => 'int2',
+ partialaggfn => 'min_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_oid', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
aggtranstype => 'oid' },
-{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
+{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
+ aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
+ aggtranstype => 'oid',
+ partialaggfn => 'min_p_oid', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_float4', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
aggtranstype => 'float4' },
-{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
+{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
+ aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
+ aggtranstype => 'float4',
+ partialaggfn => 'min_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_float8', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
aggtranstype => 'float8' },
-{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
+{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
+ aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
+ aggtranstype => 'float8',
+ partialaggfn => 'min_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_date', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
aggtranstype => 'date' },
-{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
+{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
+ aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
+ aggtranstype => 'date',
+ partialaggfn => 'min_p_date', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_time', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
aggtranstype => 'time' },
-{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
+{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
+ aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
+ aggtranstype => 'time',
+ partialaggfn => 'min_p_time', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timetz', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
aggtranstype => 'timetz' },
-{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
+{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
+ aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
+ aggtranstype => 'timetz',
+ partialaggfn => 'min_p_timetz', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_money', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
aggtranstype => 'money' },
-{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
+{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
+ aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
+ aggtranstype => 'money',
+ partialaggfn => 'min_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timestamp', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
aggtranstype => 'timestamp' },
-{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
+{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
+ aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
+ aggtranstype => 'timestamp',
+ partialaggfn => 'min_p_timestamp', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timestamptz', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
-{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
+{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
+ aggcombinefn => 'timestamptz_smaller',
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ partialaggfn => 'min_p_timestamptz', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_interval', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
aggtranstype => 'interval' },
-{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
+{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
+ aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
+ aggtranstype => 'interval',
+ partialaggfn => 'min_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_text', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
aggtranstype => 'text' },
-{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
+{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
+ aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
+ aggtranstype => 'text',
+ partialaggfn => 'min_p_text', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_numeric', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
aggtranstype => 'numeric' },
-{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
+{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
+ aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
+ aggtranstype => 'numeric',
+ partialaggfn => 'min_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_anyarray', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
aggtranstype => 'anyarray' },
-{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
+{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
+ aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
+ aggtranstype => 'anyarray',
+ partialaggfn => 'min_p_anyarray', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_bpchar', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
aggtranstype => 'bpchar' },
+{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
+ aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
+ aggtranstype => 'bpchar',
+ partialaggfn => 'min_p_bpchar', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_tid', aggtransfn => 'tidsmaller',
+ aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
+ aggtranstype => 'tid'},
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
-{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
+ aggtranstype => 'tid',
+ partialaggfn => 'min_p_tid', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_anyenum', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
aggtranstype => 'anyenum' },
-{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
+{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
+ aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
+ aggtranstype => 'anyenum',
+ partialaggfn => 'min_p_anyenum', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_inet', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
aggtranstype => 'inet' },
-{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
+{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
+ aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
+ aggtranstype => 'inet',
+ partialaggfn => 'min_p_inet', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_pg_lsn', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
aggtranstype => 'pg_lsn' },
-{ aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller',
+{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
+ aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
+ aggtranstype => 'pg_lsn',
+ partialaggfn => 'min_p_pg_lsn', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_xid8', aggtransfn => 'xid8_smaller',
aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)',
aggtranstype => 'xid8' },
+{ aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller',
+ aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)',
+ aggtranstype => 'xid8',
+ partialaggfn => 'min_p_xid8', partialagg_minversion => '160000' },
# count
+{ aggfnoid => 'count_p_any', aggtransfn => 'int8inc_any',
+ aggcombinefn => 'int8pl', aggtranstype => 'int8',
+ agginitval => '0' },
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ partialaggfn => 'count_p_any', partialagg_minversion => '160000' },
+{ aggfnoid => 'count_p', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8', agginitval => '0' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ partialaggfn => 'count_p', partialagg_minversion => '160000' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index fd2559442e..92c5bf6a82 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6490,197 +6490,389 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as bigint across all integer input values',
+ proname => 'sum_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2109', descr => 'sum as bigint across all smallint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13509', descr => 'partial sum as bigint across all smallint input values',
+ proname => 'sum_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2110', descr => 'sum as float4 across all float4 input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13510', descr => 'partial sum as float4 across all float4 input values',
+ proname => 'sum_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2111', descr => 'sum as float8 across all float8 input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13511', descr => 'partial sum as float8 across all float8 input values',
+ proname => 'sum_p_float8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2112', descr => 'sum as money across all money input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13512', descr => 'partial sum as money across all money input values',
+ proname => 'sum_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2113', descr => 'sum as interval across all interval input values',
proname => 'sum', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13513', descr => 'partial sum as interval across all interval input values',
+ proname => 'sum_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13514', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13515', descr => 'maximum value of all bigint input values',
+ proname => 'max_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2116', descr => 'maximum value of all integer input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13516', descr => 'maximum value of all integer input values',
+ proname => 'max_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2117', descr => 'maximum value of all smallint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13517', descr => 'maximum value of all smallint input values',
+ proname => 'max_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2118', descr => 'maximum value of all oid input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
proargtypes => 'oid', prosrc => 'aggregate_dummy' },
+{ oid => '13518', descr => 'maximum value of all oid input values',
+ proname => 'max_p_oid', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
+ proargtypes => 'oid', prosrc => 'aggregate_dummy' },
{ oid => '2119', descr => 'maximum value of all float4 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13519', descr => 'maximum value of all float4 input values',
+ proname => 'max_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2120', descr => 'maximum value of all float8 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13520', descr => 'maximum value of all float8 input values',
+ proname => 'max_p_floa8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2122', descr => 'maximum value of all date input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'date',
proargtypes => 'date', prosrc => 'aggregate_dummy' },
+{ oid => '13521', descr => 'maximum value of all date input values',
+ proname => 'max_p_date', prokind => 'a', proisstrict => 'f', prorettype => 'date',
+ proargtypes => 'date', prosrc => 'aggregate_dummy' },
{ oid => '2123', descr => 'maximum value of all time input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'time',
proargtypes => 'time', prosrc => 'aggregate_dummy' },
+{ oid => '13522', descr => 'maximum value of all time input values',
+ proname => 'max_p_time', prokind => 'a', proisstrict => 'f', prorettype => 'time',
+ proargtypes => 'time', prosrc => 'aggregate_dummy' },
{ oid => '2124',
descr => 'maximum value of all time with time zone input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
+{ oid => '13523',
+ descr => 'maximum value of all time with time zone input values',
+ proname => 'max_p_timetz', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
+ proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
{ oid => '2125', descr => 'maximum value of all money input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13524', descr => 'maximum value of all money input values',
+ proname => 'max_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2126', descr => 'maximum value of all timestamp input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamp', proargtypes => 'timestamp',
prosrc => 'aggregate_dummy' },
+{ oid => '13525', descr => 'maximum value of all timestamp input values',
+ proname => 'max_p_timestamp', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamp', proargtypes => 'timestamp',
+ prosrc => 'aggregate_dummy' },
{ oid => '2127',
descr => 'maximum value of all timestamp with time zone input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamptz', proargtypes => 'timestamptz',
prosrc => 'aggregate_dummy' },
+{ oid => '13526',
+ descr => 'maximum value of all timestamp with time zone input values',
+ proname => 'max_p_timestamptz', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamptz', proargtypes => 'timestamptz',
+ prosrc => 'aggregate_dummy' },
{ oid => '2128', descr => 'maximum value of all interval input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13527', descr => 'maximum value of all interval input values',
+ proname => 'max_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2129', descr => 'maximum value of all text input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'text',
proargtypes => 'text', prosrc => 'aggregate_dummy' },
+{ oid => '13528', descr => 'maximum value of all text input values',
+ proname => 'max_p_text', prokind => 'a', proisstrict => 'f', prorettype => 'text',
+ proargtypes => 'text', prosrc => 'aggregate_dummy' },
{ oid => '2130', descr => 'maximum value of all numeric input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13529', descr => 'maximum value of all numeric input values',
+ proname => 'max_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2050', descr => 'maximum value of all anyarray input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'anyarray', proargtypes => 'anyarray',
prosrc => 'aggregate_dummy' },
+{ oid => '13530', descr => 'maximum value of all anyarray input values',
+ proname => 'max_p_anyarray', prokind => 'a', proisstrict => 'f',
+ prorettype => 'anyarray', proargtypes => 'anyarray',
+ prosrc => 'aggregate_dummy' },
{ oid => '2244', descr => 'maximum value of all bpchar input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
+{ oid => '13531', descr => 'maximum value of all bpchar input values',
+ proname => 'max_p_bpchar', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
+ proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
{ oid => '2797', descr => 'maximum value of all tid input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
proargtypes => 'tid', prosrc => 'aggregate_dummy' },
+{ oid => '13532', descr => 'maximum value of all tid input values',
+ proname => 'max_p_tid', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
+ proargtypes => 'tid', prosrc => 'aggregate_dummy' },
{ oid => '3564', descr => 'maximum value of all inet input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
proargtypes => 'inet', prosrc => 'aggregate_dummy' },
+{ oid => '13533', descr => 'maximum value of all inet input values',
+ proname => 'max_p_inet', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
+ proargtypes => 'inet', prosrc => 'aggregate_dummy' },
{ oid => '4189', descr => 'maximum value of all pg_lsn input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
+{ oid => '13534', descr => 'maximum value of all pg_lsn input values',
+ proname => 'max_p_pg_lsn', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
+ proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
{ oid => '5099', descr => 'maximum value of all xid8 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
+{ oid => '13535', descr => 'maximum value of all xid8 input values',
+ proname => 'max_p_xid8', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
+ proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
{ oid => '2131', descr => 'minimum value of all bigint input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13536', descr => 'minimum value of all bigint input values',
+ proname => 'min_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2132', descr => 'minimum value of all integer input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13537', descr => 'minimum value of all integer input values',
+ proname => 'min_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2133', descr => 'minimum value of all smallint input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13538', descr => 'minimum value of all smallint input values',
+ proname => 'min_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2134', descr => 'minimum value of all oid input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
proargtypes => 'oid', prosrc => 'aggregate_dummy' },
+{ oid => '13539', descr => 'minimum value of all oid input values',
+ proname => 'min_p_oid', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
+ proargtypes => 'oid', prosrc => 'aggregate_dummy' },
{ oid => '2135', descr => 'minimum value of all float4 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13540', descr => 'minimum value of all float4 input values',
+ proname => 'min_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2136', descr => 'minimum value of all float8 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13541', descr => 'minimum value of all float8 input values',
+ proname => 'min_p_float8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2138', descr => 'minimum value of all date input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'date',
proargtypes => 'date', prosrc => 'aggregate_dummy' },
+{ oid => '13542', descr => 'minimum value of all date input values',
+ proname => 'min_p_date', prokind => 'a', proisstrict => 'f', prorettype => 'date',
+ proargtypes => 'date', prosrc => 'aggregate_dummy' },
{ oid => '2139', descr => 'minimum value of all time input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'time',
proargtypes => 'time', prosrc => 'aggregate_dummy' },
+{ oid => '13543', descr => 'minimum value of all time input values',
+ proname => 'min_p_time', prokind => 'a', proisstrict => 'f', prorettype => 'time',
+ proargtypes => 'time', prosrc => 'aggregate_dummy' },
{ oid => '2140',
descr => 'minimum value of all time with time zone input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
+{ oid => '13544',
+ descr => 'minimum value of all time with time zone input values',
+ proname => 'min_p_timetz', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
+ proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
{ oid => '2141', descr => 'minimum value of all money input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13545', descr => 'minimum value of all money input values',
+ proname => 'min_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2142', descr => 'minimum value of all timestamp input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamp', proargtypes => 'timestamp',
prosrc => 'aggregate_dummy' },
+{ oid => '13546', descr => 'minimum value of all timestamp input values',
+ proname => 'min_p_timestamp', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamp', proargtypes => 'timestamp',
+ prosrc => 'aggregate_dummy' },
{ oid => '2143',
descr => 'minimum value of all timestamp with time zone input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamptz', proargtypes => 'timestamptz',
prosrc => 'aggregate_dummy' },
+{ oid => '13547',
+ descr => 'minimum value of all timestamp with time zone input values',
+ proname => 'min_p_timestamptz', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamptz', proargtypes => 'timestamptz',
+ prosrc => 'aggregate_dummy' },
{ oid => '2144', descr => 'minimum value of all interval input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13548', descr => 'minimum value of all interval input values',
+ proname => 'min_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2145', descr => 'minimum value of all text values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'text',
proargtypes => 'text', prosrc => 'aggregate_dummy' },
+{ oid => '13549', descr => 'minimum value of all text values',
+ proname => 'min_p_text', prokind => 'a', proisstrict => 'f', prorettype => 'text',
+ proargtypes => 'text', prosrc => 'aggregate_dummy' },
{ oid => '2146', descr => 'minimum value of all numeric input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13550', descr => 'minimum value of all numeric input values',
+ proname => 'min_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2051', descr => 'minimum value of all anyarray input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'anyarray', proargtypes => 'anyarray',
prosrc => 'aggregate_dummy' },
+{ oid => '13551', descr => 'minimum value of all anyarray input values',
+ proname => 'min_p_anyarray', prokind => 'a', proisstrict => 'f',
+ prorettype => 'anyarray', proargtypes => 'anyarray',
+ prosrc => 'aggregate_dummy' },
{ oid => '2245', descr => 'minimum value of all bpchar input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
+{ oid => '13552', descr => 'minimum value of all bpchar input values',
+ proname => 'min_p_bpchar', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
+ proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
{ oid => '2798', descr => 'minimum value of all tid input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
proargtypes => 'tid', prosrc => 'aggregate_dummy' },
+{ oid => '13553', descr => 'minimum value of all tid input values',
+ proname => 'min_p_tid', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
+ proargtypes => 'tid', prosrc => 'aggregate_dummy' },
{ oid => '3565', descr => 'minimum value of all inet input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
proargtypes => 'inet', prosrc => 'aggregate_dummy' },
+{ oid => '13554', descr => 'minimum value of all inet input values',
+ proname => 'min_p_inet', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
+ proargtypes => 'inet', prosrc => 'aggregate_dummy' },
{ oid => '4190', descr => 'minimum value of all pg_lsn input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
+{ oid => '13555', descr => 'minimum value of all pg_lsn input values',
+ proname => 'min_p_pg_lsn', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
+ proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
{ oid => '5100', descr => 'minimum value of all xid8 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
+{ oid => '13556', descr => 'minimum value of all xid8 input values',
+ proname => 'min_p_xid8', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
+ proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
# count has two forms: count(any) and count(*)
{ oid => '2147',
@@ -6688,10 +6880,19 @@
proname => 'count', prosupport => 'int8inc_support', prokind => 'a',
proisstrict => 'f', prorettype => 'int8', proargtypes => 'any',
prosrc => 'aggregate_dummy' },
+{ oid => '13557',
+ descr => 'partial number of input rows for which the input expression is not null',
+ proname => 'count_p_any', prosupport => 'int8inc_support', prokind => 'a',
+ proisstrict => 'f', prorettype => 'int8', proargtypes => 'any',
+ prosrc => 'aggregate_dummy' },
{ oid => '2803', descr => 'number of input rows',
proname => 'count', prosupport => 'int8inc_support', prokind => 'a',
proisstrict => 'f', prorettype => 'int8', proargtypes => '',
prosrc => 'aggregate_dummy' },
+{ oid => '13558', descr => 'partial number of input rows',
+ proname => 'count_p', prosupport => 'int8inc_support', prokind => 'a',
+ proisstrict => 'f', prorettype => 'int8', proargtypes => '',
+ prosrc => 'aggregate_dummy' },
{ oid => '6236', descr => 'planner support for count run condition',
proname => 'int8inc_support', prorettype => 'internal',
proargtypes => 'internal', prosrc => 'int8inc_support' },
@@ -9081,9 +9282,15 @@
{ oid => '3526', descr => 'maximum value of all enum input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
+{ oid => '13559', descr => 'maximum value of all enum input values',
+ proname => 'max_p_anyenum', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
+ proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
{ oid => '3527', descr => 'minimum value of all enum input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
+{ oid => '13560', descr => 'minimum value of all enum input values',
+ proname => 'min_p_anyenum', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
+ proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
{ oid => '3528', descr => 'first value of the input enum type',
proname => 'enum_first', proisstrict => 'f', provolatile => 's',
prorettype => 'anyenum', proargtypes => 'anyenum', prosrc => 'enum_first' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/expected/opr_sanity.out b/src/test/regress/expected/opr_sanity.out
index 330eb0f765..a3ddf2ff84 100644
--- a/src/test/regress/expected/opr_sanity.out
+++ b/src/test/regress/expected/opr_sanity.out
@@ -1722,14 +1722,58 @@ SELECT DISTINCT proname, oprname
FROM pg_operator AS o, pg_aggregate AS a, pg_proc AS p
WHERE a.aggfnoid = p.oid AND a.aggsortop = o.oid
ORDER BY 1, 2;
- proname | oprname
-----------+---------
- bool_and | <
- bool_or | >
- every | <
- max | >
- min | <
-(5 rows)
+ proname | oprname
+-------------------+---------
+ bool_and | <
+ bool_or | >
+ every | <
+ max | >
+ max_p_anyarray | >
+ max_p_anyenum | >
+ max_p_bpchar | >
+ max_p_date | >
+ max_p_floa8 | >
+ max_p_float4 | >
+ max_p_inet | >
+ max_p_int2 | >
+ max_p_int4 | >
+ max_p_int8 | >
+ max_p_interval | >
+ max_p_money | >
+ max_p_numeric | >
+ max_p_oid | >
+ max_p_pg_lsn | >
+ max_p_text | >
+ max_p_tid | >
+ max_p_time | >
+ max_p_timestamp | >
+ max_p_timestamptz | >
+ max_p_timetz | >
+ max_p_xid8 | >
+ min | <
+ min_p_anyarray | <
+ min_p_anyenum | <
+ min_p_bpchar | <
+ min_p_date | <
+ min_p_float4 | <
+ min_p_float8 | <
+ min_p_inet | <
+ min_p_int2 | <
+ min_p_int4 | <
+ min_p_int8 | <
+ min_p_interval | <
+ min_p_money | <
+ min_p_numeric | <
+ min_p_oid | <
+ min_p_pg_lsn | <
+ min_p_text | <
+ min_p_tid | <
+ min_p_time | <
+ min_p_timestamp | <
+ min_p_timestamptz | <
+ min_p_timetz | <
+ min_p_xid8 | <
+(49 rows)
-- Check datatypes match
SELECT a.aggfnoid::oid, o.oid
@@ -1762,14 +1806,58 @@ WHERE a.aggfnoid = p.oid AND a.aggsortop = o.oid AND
amopopr = o.oid AND
amopmethod = (SELECT oid FROM pg_am WHERE amname = 'btree')
ORDER BY 1, 2;
- proname | oprname | amopstrategy
-----------+---------+--------------
- bool_and | < | 1
- bool_or | > | 5
- every | < | 1
- max | > | 5
- min | < | 1
-(5 rows)
+ proname | oprname | amopstrategy
+-------------------+---------+--------------
+ bool_and | < | 1
+ bool_or | > | 5
+ every | < | 1
+ max | > | 5
+ max_p_anyarray | > | 5
+ max_p_anyenum | > | 5
+ max_p_bpchar | > | 5
+ max_p_date | > | 5
+ max_p_floa8 | > | 5
+ max_p_float4 | > | 5
+ max_p_inet | > | 5
+ max_p_int2 | > | 5
+ max_p_int4 | > | 5
+ max_p_int8 | > | 5
+ max_p_interval | > | 5
+ max_p_money | > | 5
+ max_p_numeric | > | 5
+ max_p_oid | > | 5
+ max_p_pg_lsn | > | 5
+ max_p_text | > | 5
+ max_p_tid | > | 5
+ max_p_time | > | 5
+ max_p_timestamp | > | 5
+ max_p_timestamptz | > | 5
+ max_p_timetz | > | 5
+ max_p_xid8 | > | 5
+ min | < | 1
+ min_p_anyarray | < | 1
+ min_p_anyenum | < | 1
+ min_p_bpchar | < | 1
+ min_p_date | < | 1
+ min_p_float4 | < | 1
+ min_p_float8 | < | 1
+ min_p_inet | < | 1
+ min_p_int2 | < | 1
+ min_p_int4 | < | 1
+ min_p_int8 | < | 1
+ min_p_interval | < | 1
+ min_p_money | < | 1
+ min_p_numeric | < | 1
+ min_p_oid | < | 1
+ min_p_pg_lsn | < | 1
+ min_p_text | < | 1
+ min_p_tid | < | 1
+ min_p_time | < | 1
+ min_p_timestamp | < | 1
+ min_p_timestamptz | < | 1
+ min_p_timetz | < | 1
+ min_p_xid8 | < | 1
+(49 rows)
-- Check that there are not aggregates with the same name and different
-- numbers of arguments. While not technically wrong, we have a project policy
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-22 05:00 ` Ted Yu <[email protected]>
2022-11-22 09:11 ` RE: [CAUTION!! freemail] Re: Partial aggregates pushdown [email protected] <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Ted Yu @ 2022-11-22 05:00 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
On Mon, Nov 21, 2022 at 5:02 PM [email protected] <
[email protected]> wrote:
> Hi Mr.Vondra, Mr.Pyhalov, Everyone.
>
> I discussed with Mr.Pyhalov about the above draft by directly sending mail
> to
> him(outside of pgsql-hackers). Mr.Pyhalov allowed me to update his patch
> along with the above draft. So I update Mr.Pyhalov's patch v10.
>
> I wrote my patch for discussion.
> My patch passes regression tests which contains additional basic
> postgres_fdw tests
> for my patch's feature. But my patch doesn't contain sufficient documents
> and tests.
> If reviewers accept my approach, I will add documents and tests to my
> patch.
>
> The following is a my patch's readme.
> # I simplified the above draft.
>
> --readme of my patch
> 1. interface
> 1) pg_aggregate
> There are the following additional columns.
> a) partialaggfn
> data type : regproc.
> default value: zero(means invalid).
> description : This field refers to the special aggregate function(then
> we call
> this partialaggfunc)
> corresponding to aggregation function(then we call src) which has
> aggfnoid.
> partialaggfunc is used for partial aggregation pushdown by
> postgres_fdw.
> The followings are differences between the src and the special
> aggregate function.
> difference1) result type
> The result type is same as the src's transtype if the src's
> transtype
> is not internal.
> Otherwise the result type is bytea.
> difference2) final func
> The final func does not exist if the src's transtype is not
> internal.
> Otherwize the final func returns serialized value.
> For example, there is a partialaggfunc avg_p_int4 which corresponds to
> avg(int4)
> whose aggtranstype is _int4.
> The result value of avg_p_int4 is a float8 array which consists of
> count and
> summation. avg_p_int4 does not have finalfunc.
> For another example, there is a partialaggfunc avg_p_int8 which
> corresponds to
> avg(int8) whose aggtranstype is internal.
> The result value of avg_p_int8 is a bytea serialized array which
> consists of count
> and summation. avg_p_int8 has finalfunc int8_avg_serialize which is
> serialize function
> of avg(int8). This field is zero if there is no partialaggfunc.
>
> b) partialagg_minversion
> data type : int4.
> default value: zero(means current version).
> description : This field is the minimum PostgreSQL server version which
> has
> partialaggfunc. This field is used for checking compatibility of
> partialaggfunc.
>
> The above fields are valid in tuples for builtin avg, sum, min, max, count.
> There are additional records which correspond to partialaggfunc for avg,
> sum, min, max,
> count.
>
> 2) pg_proc
> There are additional records which correspond to partialaggfunc for avg,
> sum, min, max,
> count.
>
> 3) postgres_fdw
> postgres_fdw has an additional foreign server option server_version.
> server_version is
> integer value which means remote server version number. Default value of
> server_version
> is zero. server_version is used for checking compatibility of
> partialaggfunc.
>
> 2. feature
> postgres_fdw can pushdown partial aggregation of avg, sum, min, max, count.
> Partial aggregation pushdown is fine when the following two conditions are
> both true.
> condition1) partialaggfn is valid.
> condition2) server_version is not less than partialagg_minversion
> postgres_fdw executes pushdown the patialaggfunc instead of a src.
> For example, we issue "select avg_p_int4(c) from t" instead of "select
> avg(c) from t"
> in the above example.
>
> postgres_fdw can pushdown every aggregate function which supports partial
> aggregation
> if you add a partialaggfunc corresponding to the aggregate function by
> create aggregate
> command.
>
> 3. difference between my patch and Mr.Pyhalov's v10 patch.
> 1) In my patch postgres_fdw can pushdown partial aggregation of avg
> 2) In my patch postgres_fdw can pushdown every aggregate function which
> supports partial
> aggregation if you add a partialaggfunc corresponding to the aggregate
> function.
>
> 4. sample commands in psql
> \c postgres
> drop database tmp;
> create database tmp;
> \c tmp
> create extension postgres_fdw;
> create server server_01 foreign data wrapper postgres_fdw options(host
> 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
> create user mapping for postgres server server_01 options(user 'postgres',
> password 'postgres');
> create server server_02 foreign data wrapper postgres_fdw options(host
> 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
> create user mapping for postgres server server_02 options(user 'postgres',
> password 'postgres');
>
> create table t(dt timestamp, id int4, name text, total int4, val float4,
> type int4, span interval) partition by list (type);
>
> create table t1(dt timestamp, id int4, name text, total int4, val float4,
> type int4, span interval);
> create table t2(dt timestamp, id int4, name text, total int4, val float4,
> type int4, span interval);
>
> truncate table t1;
> truncate table t2;
> insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 1, 1.1, 1, cast('1 seconds' as interval)
> from generate_series(1, 100000, 1) t;
> insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 2, 2.1, 1, cast('2 seconds' as interval)
> from generate_series(1, 100000, 1) t;
> insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 1, 1.1, 2, cast('1 seconds' as interval)
> from generate_series(1, 100000, 1) t;
> insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 2, 2.1, 2, cast('2 seconds' as interval)
> from generate_series(1, 100000, 1) t;
>
> create foreign table f_t1 partition of t for values in (1) server
> server_01 options(table_name 't1');
> create foreign table f_t2 partition of t for values in (2) server
> server_02 options(table_name 't2');
>
> set enable_partitionwise_aggregate = on;
> explain (verbose, costs off) select avg(total::int4), avg(total::int8)
> from t;
> select avg(total::int4), avg(total::int8) from t;
>
> Sincerely yours,
> Yuuki Fujii
> --
> Yuuki Fujii
> Information Technology R&D Center Mitsubishi Electric Corporation
>
Hi,
For partial_agg_compatible :
+ * Check that partial aggregate agg has compatibility
If the `agg` refers to func parameter, the parameter name is aggform
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion ==
PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = PG_VERSION_NUM;
I am curious why the same variable is assigned the same value twice. It
seems the if block is redundant.
+ if ((fpinfo->server_version >= partialagg_minversion)) {
+ compatible = true;
The above can be simplified as: return fpinfo->server_version >=
partialagg_minversion;
Cheers
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: [CAUTION!! freemail] Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 05:00 ` Re: Partial aggregates pushdown Ted Yu <[email protected]>
@ 2022-11-22 09:11 ` [email protected] <[email protected]>
2022-11-22 10:51 ` Re: [CAUTION!! freemail] Re: Partial aggregates pushdown Ted Yu <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2022-11-22 09:11 UTC (permalink / raw)
To: Ted Yu <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Yu.
Thank you for comments.
> + * Check that partial aggregate agg has compatibility
>
> If the `agg` refers to func parameter, the parameter name is aggform
I fixed the above typo and made the above comment easy to understand
New comment is "Check that partial aggregate function of aggform exsits in remote"
> + int32 partialagg_minversion = PG_VERSION_NUM;
> + if (aggform->partialagg_minversion ==
> PARTIALAGG_MINVERSION_DEFAULT) {
> + partialagg_minversion = PG_VERSION_NUM;
>
>
> I am curious why the same variable is assigned the same value twice. It seems
> the if block is redundant.
>
> + if ((fpinfo->server_version >= partialagg_minversion)) {
> + compatible = true;
>
>
> The above can be simplified as: return fpinfo->server_version >=
> partialagg_minversion;
I fixed according to your comment.
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
> -----Original Message-----
> From: Ted Yu <[email protected]>
> Sent: Tuesday, November 22, 2022 2:00 PM
> To: Fujii Yuki/藤井 雄規(MELCO/情報総研 ＤＭ最適Ｇ)
> <[email protected]>
> Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra
> <[email protected]>; PostgreSQL-development
> <[email protected]>; Andres Freund <[email protected]>;
> Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>;
> Daniel Gustafsson <[email protected]>; Ilya Gladyshev
> <[email protected]>
> Subject: [CAUTION!! freemail] Re: Partial aggregates pushdown
>
>
>
> On Mon, Nov 21, 2022 at 5:02 PM [email protected]
> <mailto:[email protected]>
> <[email protected]
> <mailto:[email protected]> > wrote:
>
>
> Hi Mr.Vondra, Mr.Pyhalov, Everyone.
>
> I discussed with Mr.Pyhalov about the above draft by directly sending
> mail to
> him(outside of pgsql-hackers). Mr.Pyhalov allowed me to update his
> patch
> along with the above draft. So I update Mr.Pyhalov's patch v10.
>
> I wrote my patch for discussion.
> My patch passes regression tests which contains additional basic
> postgres_fdw tests
> for my patch's feature. But my patch doesn't contain sufficient
> documents and tests.
> If reviewers accept my approach, I will add documents and tests to my
> patch.
>
> The following is a my patch's readme.
> # I simplified the above draft.
>
> --readme of my patch
> 1. interface
> 1) pg_aggregate
> There are the following additional columns.
> a) partialaggfn
> data type : regproc.
> default value: zero(means invalid).
> description : This field refers to the special aggregate
> function(then we call
> this partialaggfunc)
> corresponding to aggregation function(then we call src) which has
> aggfnoid.
> partialaggfunc is used for partial aggregation pushdown by
> postgres_fdw.
> The followings are differences between the src and the special
> aggregate function.
> difference1) result type
> The result type is same as the src's transtype if the src's
> transtype
> is not internal.
> Otherwise the result type is bytea.
> difference2) final func
> The final func does not exist if the src's transtype is not
> internal.
> Otherwize the final func returns serialized value.
> For example, there is a partialaggfunc avg_p_int4 which
> corresponds to avg(int4)
> whose aggtranstype is _int4.
> The result value of avg_p_int4 is a float8 array which consists of
> count and
> summation. avg_p_int4 does not have finalfunc.
> For another example, there is a partialaggfunc avg_p_int8 which
> corresponds to
> avg(int8) whose aggtranstype is internal.
> The result value of avg_p_int8 is a bytea serialized array which
> consists of count
> and summation. avg_p_int8 has finalfunc int8_avg_serialize
> which is serialize function
> of avg(int8). This field is zero if there is no partialaggfunc.
>
> b) partialagg_minversion
> data type : int4.
> default value: zero(means current version).
> description : This field is the minimum PostgreSQL server version
> which has
> partialaggfunc. This field is used for checking compatibility of
> partialaggfunc.
>
> The above fields are valid in tuples for builtin avg, sum, min, max,
> count.
> There are additional records which correspond to partialaggfunc for
> avg, sum, min, max,
> count.
>
> 2) pg_proc
> There are additional records which correspond to partialaggfunc for
> avg, sum, min, max,
> count.
>
> 3) postgres_fdw
> postgres_fdw has an additional foreign server option server_version.
> server_version is
> integer value which means remote server version number. Default
> value of server_version
> is zero. server_version is used for checking compatibility of
> partialaggfunc.
>
> 2. feature
> postgres_fdw can pushdown partial aggregation of avg, sum, min, max,
> count.
> Partial aggregation pushdown is fine when the following two
> conditions are both true.
> condition1) partialaggfn is valid.
> condition2) server_version is not less than partialagg_minversion
> postgres_fdw executes pushdown the patialaggfunc instead of a src.
> For example, we issue "select avg_p_int4(c) from t" instead of "select
> avg(c) from t"
> in the above example.
>
> postgres_fdw can pushdown every aggregate function which supports
> partial aggregation
> if you add a partialaggfunc corresponding to the aggregate function by
> create aggregate
> command.
>
> 3. difference between my patch and Mr.Pyhalov's v10 patch.
> 1) In my patch postgres_fdw can pushdown partial aggregation of avg
> 2) In my patch postgres_fdw can pushdown every aggregate function
> which supports partial
> aggregation if you add a partialaggfunc corresponding to the
> aggregate function.
>
> 4. sample commands in psql
> \c postgres
> drop database tmp;
> create database tmp;
> \c tmp
> create extension postgres_fdw;
> create server server_01 foreign data wrapper postgres_fdw
> options(host 'localhost', dbname 'tmp', server_version '160000', async_capable
> 'true');
> create user mapping for postgres server server_01 options(user
> 'postgres', password 'postgres');
> create server server_02 foreign data wrapper postgres_fdw
> options(host 'localhost', dbname 'tmp', server_version '160000', async_capable
> 'true');
> create user mapping for postgres server server_02 options(user
> 'postgres', password 'postgres');
>
> create table t(dt timestamp, id int4, name text, total int4, val float4, type
> int4, span interval) partition by list (type);
>
> create table t1(dt timestamp, id int4, name text, total int4, val float4,
> type int4, span interval);
> create table t2(dt timestamp, id int4, name text, total int4, val float4,
> type int4, span interval);
>
> truncate table t1;
> truncate table t2;
> insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 1, 1.1, 1, cast('1 seconds' as interval) from
> generate_series(1, 100000, 1) t;
> insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 2, 2.1, 1, cast('2 seconds' as interval) from
> generate_series(1, 100000, 1) t;
> insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 1, 1.1, 2, cast('1 seconds' as interval) from
> generate_series(1, 100000, 1) t;
> insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as
> interval), t % 100, 'hoge' || t, 2, 2.1, 2, cast('2 seconds' as interval) from
> generate_series(1, 100000, 1) t;
>
> create foreign table f_t1 partition of t for values in (1) server server_01
> options(table_name 't1');
> create foreign table f_t2 partition of t for values in (2) server server_02
> options(table_name 't2');
>
> set enable_partitionwise_aggregate = on;
> explain (verbose, costs off) select avg(total::int4), avg(total::int8) from
> t;
> select avg(total::int4), avg(total::int8) from t;
>
> Sincerely yours,
> Yuuki Fujii
> --
> Yuuki Fujii
> Information Technology R&D Center Mitsubishi Electric Corporation
>
>
>
> Hi,
> For partial_agg_compatible :
>
> + * Check that partial aggregate agg has compatibility
>
> If the `agg` refers to func parameter, the parameter name is aggform
>
> + int32 partialagg_minversion = PG_VERSION_NUM;
> + if (aggform->partialagg_minversion ==
> PARTIALAGG_MINVERSION_DEFAULT) {
> + partialagg_minversion = PG_VERSION_NUM;
>
>
> I am curious why the same variable is assigned the same value twice. It seems
> the if block is redundant.
>
> + if ((fpinfo->server_version >= partialagg_minversion)) {
> + compatible = true;
>
>
> The above can be simplified as: return fpinfo->server_version >=
> partialagg_minversion;
>
> Cheers
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v12.patch (106.9K, ../../OS3PR01MB6660AE8BB4D08125EB4D3279950D9@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v12.patch)
download | inline diff:
From bd59c95796dde0b063fb7d96591740dbcdc76534 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Tue, 22 Nov 2022 17:00:01 +0900
Subject: [PATCH] Partial aggregates push down v12
---
contrib/postgres_fdw/deparse.c | 84 +++-
.../postgres_fdw/expected/postgres_fdw.out | 246 +++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 25 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 54 ++-
src/backend/catalog/pg_aggregate.c | 50 ++-
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 367 +++++++++++++++---
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 207 ++++++++++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/expected/opr_sanity.out | 120 +++++-
14 files changed, 1110 insertions(+), 115 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..f48d57bd12 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,33 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+ if (aggform->partialaggfn) {
+ appendFunctionName((Oid)(aggform->partialaggfn), context);
+ } else {
+ elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3981,53 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function of aggform exsits in remote
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ *
+ * Check that it's AGGKIND_NORMAL aggregate without
+ * distinct or order by clauses, which has valid partialaggfn
+ * and partialagg_minversion <= server_version
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn)
+ {
+ partial_agg_ok = false;
+ } else if(!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..cf0f270946 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS name AS $$
+ SELECT (setting::int4 + version_offset)::name FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name name, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num name;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,213 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | avg | avg
+----+-----+-----+-------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min_p_int4(a), max_p_int4(a), count_p(*), sum_p_int4(d), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | avg | avg
+-----+-----+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(name, text, int4);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..c8efdfefb7 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -527,7 +527,7 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra, bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -6159,7 +6162,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6176,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6416,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6423,7 +6431,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6448,7 +6460,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra, bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6477,6 +6489,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
fpinfo->table = ifpinfo->table;
fpinfo->server = ifpinfo->server;
fpinfo->user = ifpinfo->user;
+ fpinfo->server_version = ifpinfo->server_version;
merge_fdw_options(fpinfo, ifpinfo, NULL);
/*
@@ -6485,7 +6498,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..5bb35d9743 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for check compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..9176337420 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS name AS $$
+ SELECT (setting::int4 + version_offset)::name FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name name, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num name;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,34 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(name, text, int4);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..9f8a326cec 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -36,7 +36,8 @@
static Oid lookup_agg_function(List *fnName, int nargs, Oid *input_types,
Oid variadicArgType,
- Oid *rettype);
+ Oid *rettype,
+ bool only_normal);
/*
@@ -63,6 +64,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +74,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +94,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -228,7 +232,7 @@ AggregateCreate(const char *aggName,
}
transfn = lookup_agg_function(aggtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/*
* Return type of transfn (possibly after refinement by
@@ -281,7 +285,7 @@ AggregateCreate(const char *aggName,
mtransfn = lookup_agg_function(aggmtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/* As above, return type must exactly match declared mtranstype. */
if (rettype != aggmTransType)
@@ -326,7 +330,7 @@ AggregateCreate(const char *aggName,
minvtransfn = lookup_agg_function(aggminvtransfnName, nargs_transfn,
fnArgs, variadicArgType,
- &rettype);
+ &rettype, true);
/* As above, return type must exactly match declared mtranstype. */
if (rettype != aggmTransType)
@@ -382,7 +386,7 @@ AggregateCreate(const char *aggName,
finalfn = lookup_agg_function(aggfinalfnName, nargs_finalfn,
fnArgs, ffnVariadicArgType,
- &finaltype);
+ &finaltype, true);
/*
* When finalfnExtraArgs is specified, the finalfn will certainly be
@@ -418,7 +422,7 @@ AggregateCreate(const char *aggName,
combinefn = lookup_agg_function(aggcombinefnName, 2,
fnArgs, InvalidOid,
- &combineType);
+ &combineType, true);
/* Ensure the return type matches the aggregate's trans type */
if (combineType != aggTransType)
@@ -450,7 +454,7 @@ AggregateCreate(const char *aggName,
serialfn = lookup_agg_function(aggserialfnName, 1,
fnArgs, InvalidOid,
- &rettype);
+ &rettype, true);
if (rettype != BYTEAOID)
ereport(ERROR,
@@ -471,7 +475,7 @@ AggregateCreate(const char *aggName,
deserialfn = lookup_agg_function(aggdeserialfnName, 2,
fnArgs, InvalidOid,
- &rettype);
+ &rettype, true);
if (rettype != INTERNALOID)
ereport(ERROR,
@@ -545,7 +549,7 @@ AggregateCreate(const char *aggName,
mfinalfn = lookup_agg_function(aggmfinalfnName, nargs_finalfn,
fnArgs, ffnVariadicArgType,
- &rettype);
+ &rettype, true);
/* As above, check strictness if mfinalfnExtraArgs is given */
if (mfinalfnExtraArgs && func_strict(mfinalfn))
@@ -569,6 +573,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = lookup_agg_function(partialaggfnName, numArgs,
+ aggArgTypes, variadicArgType, &rettype, false);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial aggregation function %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +709,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
@@ -827,7 +854,8 @@ lookup_agg_function(List *fnName,
int nargs,
Oid *input_types,
Oid variadicArgType,
- Oid *rettype)
+ Oid *rettype,
+ bool only_normal)
{
Oid fnOid;
bool retset;
@@ -852,7 +880,7 @@ lookup_agg_function(List *fnName,
&true_oid_array, NULL);
/* only valid case is a normal function not returning a set */
- if (fdresult != FUNCDETAIL_NORMAL || !OidIsValid(fnOid))
+ if ((fdresult != FUNCDETAIL_NORMAL || !OidIsValid(fnOid)) && only_normal)
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_FUNCTION),
errmsg("function %s does not exist",
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..fb1cceeb62 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,46 +58,87 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_int4', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}',
+ partialaggfn => 'sum_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_int2', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
-{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
+ aggmtranstype => '_int8', aggminitval => '{0,0}',
+ partialaggfn => 'sum_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_float4', aggtransfn => 'float4pl',
aggcombinefn => 'float4pl', aggtranstype => 'float4' },
-{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
+{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
+ aggcombinefn => 'float4pl', aggtranstype => 'float4',
+ partialaggfn => 'sum_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_float8', aggtransfn => 'float8pl',
aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
+ aggcombinefn => 'float8pl', aggtranstype => 'float8',
+ partialaggfn => 'sum_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_money', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
+ aggtranstype => 'money' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money',
+ partialaggfn => 'sum_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_interval', aggtransfn => 'interval_pl',
+ aggcombinefn => 'interval_pl', aggtranstype => 'interval' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval',
+ partialaggfn => 'sum_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,152 +146,336 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
-{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
+{ aggfnoid => 'max_p_int8', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
aggtranstype => 'int8' },
-{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
+{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
+ aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
+ aggtranstype => 'int8',
+ partialaggfn => 'max_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_int4', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
aggtranstype => 'int4' },
-{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
+{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
+ aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
+ aggtranstype => 'int4',
+ partialaggfn => 'max_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_int2', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
aggtranstype => 'int2' },
-{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
+{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
+ aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
+ aggtranstype => 'int2',
+ partialaggfn => 'max_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_oid', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
aggtranstype => 'oid' },
-{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
+{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
+ aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
+ aggtranstype => 'oid',
+ partialaggfn => 'max_p_oid', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_float4', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
aggtranstype => 'float4' },
-{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
+{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
+ aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
+ aggtranstype => 'float4',
+ partialaggfn => 'max_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_floa8', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
aggtranstype => 'float8' },
-{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
+{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
+ aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
+ aggtranstype => 'float8',
+ partialaggfn => 'max_p_floa8', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_date', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
aggtranstype => 'date' },
-{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
+{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
+ aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
+ aggtranstype => 'date',
+ partialaggfn => 'max_p_date', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_time', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
aggtranstype => 'time' },
-{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
+{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
+ aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
+ aggtranstype => 'time',
+ partialaggfn => 'max_p_time', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timetz', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
aggtranstype => 'timetz' },
-{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
+{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
+ aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
+ aggtranstype => 'timetz',
+ partialaggfn => 'max_p_timetz', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_money', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
aggtranstype => 'money' },
-{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
+{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
+ aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
+ aggtranstype => 'money',
+ partialaggfn => 'max_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timestamp', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
aggtranstype => 'timestamp' },
-{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
+{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
+ aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
+ aggtranstype => 'timestamp',
+ partialaggfn => 'max_p_timestamp', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_timestamptz', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
-{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
+{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
+ aggcombinefn => 'timestamptz_larger',
+ aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ partialaggfn => 'max_p_timestamptz', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_interval', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
aggtranstype => 'interval' },
-{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
+{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
+ aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
+ aggtranstype => 'interval',
+ partialaggfn => 'max_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_text', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
aggtranstype => 'text' },
-{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
+{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
+ aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
+ aggtranstype => 'text',
+ partialaggfn => 'max_p_text', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_numeric', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
aggtranstype => 'numeric' },
-{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
+{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
+ aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
+ aggtranstype => 'numeric',
+ partialaggfn => 'max_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_anyarray', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
aggtranstype => 'anyarray' },
-{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
+{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
+ aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
+ aggtranstype => 'anyarray',
+ partialaggfn => 'max_p_anyarray', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_bpchar', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
aggtranstype => 'bpchar' },
-{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
+{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
+ aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
+ aggtranstype => 'bpchar',
+ partialaggfn => 'max_p_bpchar', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_tid', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
aggtranstype => 'tid' },
-{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
+{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
+ aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
+ aggtranstype => 'tid',
+ partialaggfn => 'max_p_tid', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_anyenum', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
aggtranstype => 'anyenum' },
-{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
+{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
+ aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
+ aggtranstype => 'anyenum',
+ partialaggfn => 'max_p_anyenum', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_inet', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
aggtranstype => 'inet' },
-{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
+{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
+ aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
+ aggtranstype => 'inet',
+ partialaggfn => 'max_p_inet', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_pg_lsn', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
aggtranstype => 'pg_lsn' },
-{ aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger',
+{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
+ aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
+ aggtranstype => 'pg_lsn',
+ partialaggfn => 'max_p_pg_lsn', partialagg_minversion => '160000' },
+{ aggfnoid => 'max_p_xid8', aggtransfn => 'xid8_larger',
aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)',
aggtranstype => 'xid8' },
+{ aggfnoid => 'max(xid8)', aggtransfn => 'xid8_larger',
+ aggcombinefn => 'xid8_larger', aggsortop => '>(xid8,xid8)',
+ aggtranstype => 'xid8',
+ partialaggfn => 'max_p_xid8', partialagg_minversion => '160000' },
# min
-{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
+{ aggfnoid => 'min_p_int8', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
aggtranstype => 'int8' },
-{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
+{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
+ aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
+ aggtranstype => 'int8',
+ partialaggfn => 'min_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_int4', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
aggtranstype => 'int4' },
-{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
+{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
+ aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
+ aggtranstype => 'int4',
+ partialaggfn => 'min_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_int2', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
aggtranstype => 'int2' },
-{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
+{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
+ aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
+ aggtranstype => 'int2',
+ partialaggfn => 'min_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_oid', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
aggtranstype => 'oid' },
-{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
+{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
+ aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
+ aggtranstype => 'oid',
+ partialaggfn => 'min_p_oid', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_float4', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
aggtranstype => 'float4' },
-{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
+{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
+ aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
+ aggtranstype => 'float4',
+ partialaggfn => 'min_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_float8', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
aggtranstype => 'float8' },
-{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
+{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
+ aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
+ aggtranstype => 'float8',
+ partialaggfn => 'min_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_date', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
aggtranstype => 'date' },
-{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
+{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
+ aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
+ aggtranstype => 'date',
+ partialaggfn => 'min_p_date', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_time', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
aggtranstype => 'time' },
-{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
+{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
+ aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
+ aggtranstype => 'time',
+ partialaggfn => 'min_p_time', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timetz', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
aggtranstype => 'timetz' },
-{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
+{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
+ aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
+ aggtranstype => 'timetz',
+ partialaggfn => 'min_p_timetz', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_money', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
aggtranstype => 'money' },
-{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
+{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
+ aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
+ aggtranstype => 'money',
+ partialaggfn => 'min_p_money', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timestamp', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
aggtranstype => 'timestamp' },
-{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
+{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
+ aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
+ aggtranstype => 'timestamp',
+ partialaggfn => 'min_p_timestamp', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_timestamptz', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
-{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
+{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
+ aggcombinefn => 'timestamptz_smaller',
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ partialaggfn => 'min_p_timestamptz', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_interval', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
aggtranstype => 'interval' },
-{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
+{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
+ aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
+ aggtranstype => 'interval',
+ partialaggfn => 'min_p_interval', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_text', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
aggtranstype => 'text' },
-{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
+{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
+ aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
+ aggtranstype => 'text',
+ partialaggfn => 'min_p_text', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_numeric', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
aggtranstype => 'numeric' },
-{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
+{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
+ aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
+ aggtranstype => 'numeric',
+ partialaggfn => 'min_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_anyarray', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
aggtranstype => 'anyarray' },
-{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
+{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
+ aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
+ aggtranstype => 'anyarray',
+ partialaggfn => 'min_p_anyarray', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_bpchar', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
aggtranstype => 'bpchar' },
+{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
+ aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
+ aggtranstype => 'bpchar',
+ partialaggfn => 'min_p_bpchar', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_tid', aggtransfn => 'tidsmaller',
+ aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
+ aggtranstype => 'tid'},
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
-{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
+ aggtranstype => 'tid',
+ partialaggfn => 'min_p_tid', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_anyenum', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
aggtranstype => 'anyenum' },
-{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
+{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
+ aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
+ aggtranstype => 'anyenum',
+ partialaggfn => 'min_p_anyenum', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_inet', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
aggtranstype => 'inet' },
-{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
+{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
+ aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
+ aggtranstype => 'inet',
+ partialaggfn => 'min_p_inet', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_pg_lsn', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
aggtranstype => 'pg_lsn' },
-{ aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller',
+{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
+ aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
+ aggtranstype => 'pg_lsn',
+ partialaggfn => 'min_p_pg_lsn', partialagg_minversion => '160000' },
+{ aggfnoid => 'min_p_xid8', aggtransfn => 'xid8_smaller',
aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)',
aggtranstype => 'xid8' },
+{ aggfnoid => 'min(xid8)', aggtransfn => 'xid8_smaller',
+ aggcombinefn => 'xid8_smaller', aggsortop => '<(xid8,xid8)',
+ aggtranstype => 'xid8',
+ partialaggfn => 'min_p_xid8', partialagg_minversion => '160000' },
# count
+{ aggfnoid => 'count_p_any', aggtransfn => 'int8inc_any',
+ aggcombinefn => 'int8pl', aggtranstype => 'int8',
+ agginitval => '0' },
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ partialaggfn => 'count_p_any', partialagg_minversion => '160000' },
+{ aggfnoid => 'count_p', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
+ aggtranstype => 'int8', agginitval => '0' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ partialaggfn => 'count_p', partialagg_minversion => '160000' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index fd2559442e..92c5bf6a82 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6490,197 +6490,389 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as bigint across all integer input values',
+ proname => 'sum_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2109', descr => 'sum as bigint across all smallint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13509', descr => 'partial sum as bigint across all smallint input values',
+ proname => 'sum_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2110', descr => 'sum as float4 across all float4 input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13510', descr => 'partial sum as float4 across all float4 input values',
+ proname => 'sum_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2111', descr => 'sum as float8 across all float8 input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13511', descr => 'partial sum as float8 across all float8 input values',
+ proname => 'sum_p_float8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2112', descr => 'sum as money across all money input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13512', descr => 'partial sum as money across all money input values',
+ proname => 'sum_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2113', descr => 'sum as interval across all interval input values',
proname => 'sum', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13513', descr => 'partial sum as interval across all interval input values',
+ proname => 'sum_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13514', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13515', descr => 'maximum value of all bigint input values',
+ proname => 'max_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2116', descr => 'maximum value of all integer input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13516', descr => 'maximum value of all integer input values',
+ proname => 'max_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2117', descr => 'maximum value of all smallint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13517', descr => 'maximum value of all smallint input values',
+ proname => 'max_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2118', descr => 'maximum value of all oid input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
proargtypes => 'oid', prosrc => 'aggregate_dummy' },
+{ oid => '13518', descr => 'maximum value of all oid input values',
+ proname => 'max_p_oid', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
+ proargtypes => 'oid', prosrc => 'aggregate_dummy' },
{ oid => '2119', descr => 'maximum value of all float4 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13519', descr => 'maximum value of all float4 input values',
+ proname => 'max_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2120', descr => 'maximum value of all float8 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13520', descr => 'maximum value of all float8 input values',
+ proname => 'max_p_floa8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2122', descr => 'maximum value of all date input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'date',
proargtypes => 'date', prosrc => 'aggregate_dummy' },
+{ oid => '13521', descr => 'maximum value of all date input values',
+ proname => 'max_p_date', prokind => 'a', proisstrict => 'f', prorettype => 'date',
+ proargtypes => 'date', prosrc => 'aggregate_dummy' },
{ oid => '2123', descr => 'maximum value of all time input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'time',
proargtypes => 'time', prosrc => 'aggregate_dummy' },
+{ oid => '13522', descr => 'maximum value of all time input values',
+ proname => 'max_p_time', prokind => 'a', proisstrict => 'f', prorettype => 'time',
+ proargtypes => 'time', prosrc => 'aggregate_dummy' },
{ oid => '2124',
descr => 'maximum value of all time with time zone input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
+{ oid => '13523',
+ descr => 'maximum value of all time with time zone input values',
+ proname => 'max_p_timetz', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
+ proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
{ oid => '2125', descr => 'maximum value of all money input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13524', descr => 'maximum value of all money input values',
+ proname => 'max_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2126', descr => 'maximum value of all timestamp input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamp', proargtypes => 'timestamp',
prosrc => 'aggregate_dummy' },
+{ oid => '13525', descr => 'maximum value of all timestamp input values',
+ proname => 'max_p_timestamp', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamp', proargtypes => 'timestamp',
+ prosrc => 'aggregate_dummy' },
{ oid => '2127',
descr => 'maximum value of all timestamp with time zone input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamptz', proargtypes => 'timestamptz',
prosrc => 'aggregate_dummy' },
+{ oid => '13526',
+ descr => 'maximum value of all timestamp with time zone input values',
+ proname => 'max_p_timestamptz', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamptz', proargtypes => 'timestamptz',
+ prosrc => 'aggregate_dummy' },
{ oid => '2128', descr => 'maximum value of all interval input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13527', descr => 'maximum value of all interval input values',
+ proname => 'max_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2129', descr => 'maximum value of all text input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'text',
proargtypes => 'text', prosrc => 'aggregate_dummy' },
+{ oid => '13528', descr => 'maximum value of all text input values',
+ proname => 'max_p_text', prokind => 'a', proisstrict => 'f', prorettype => 'text',
+ proargtypes => 'text', prosrc => 'aggregate_dummy' },
{ oid => '2130', descr => 'maximum value of all numeric input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13529', descr => 'maximum value of all numeric input values',
+ proname => 'max_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2050', descr => 'maximum value of all anyarray input values',
proname => 'max', prokind => 'a', proisstrict => 'f',
prorettype => 'anyarray', proargtypes => 'anyarray',
prosrc => 'aggregate_dummy' },
+{ oid => '13530', descr => 'maximum value of all anyarray input values',
+ proname => 'max_p_anyarray', prokind => 'a', proisstrict => 'f',
+ prorettype => 'anyarray', proargtypes => 'anyarray',
+ prosrc => 'aggregate_dummy' },
{ oid => '2244', descr => 'maximum value of all bpchar input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
+{ oid => '13531', descr => 'maximum value of all bpchar input values',
+ proname => 'max_p_bpchar', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
+ proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
{ oid => '2797', descr => 'maximum value of all tid input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
proargtypes => 'tid', prosrc => 'aggregate_dummy' },
+{ oid => '13532', descr => 'maximum value of all tid input values',
+ proname => 'max_p_tid', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
+ proargtypes => 'tid', prosrc => 'aggregate_dummy' },
{ oid => '3564', descr => 'maximum value of all inet input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
proargtypes => 'inet', prosrc => 'aggregate_dummy' },
+{ oid => '13533', descr => 'maximum value of all inet input values',
+ proname => 'max_p_inet', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
+ proargtypes => 'inet', prosrc => 'aggregate_dummy' },
{ oid => '4189', descr => 'maximum value of all pg_lsn input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
+{ oid => '13534', descr => 'maximum value of all pg_lsn input values',
+ proname => 'max_p_pg_lsn', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
+ proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
{ oid => '5099', descr => 'maximum value of all xid8 input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
+{ oid => '13535', descr => 'maximum value of all xid8 input values',
+ proname => 'max_p_xid8', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
+ proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
{ oid => '2131', descr => 'minimum value of all bigint input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13536', descr => 'minimum value of all bigint input values',
+ proname => 'min_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2132', descr => 'minimum value of all integer input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13537', descr => 'minimum value of all integer input values',
+ proname => 'min_p_int4', prokind => 'a', proisstrict => 'f', prorettype => 'int4',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2133', descr => 'minimum value of all smallint input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13538', descr => 'minimum value of all smallint input values',
+ proname => 'min_p_int2', prokind => 'a', proisstrict => 'f', prorettype => 'int2',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2134', descr => 'minimum value of all oid input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
proargtypes => 'oid', prosrc => 'aggregate_dummy' },
+{ oid => '13539', descr => 'minimum value of all oid input values',
+ proname => 'min_p_oid', prokind => 'a', proisstrict => 'f', prorettype => 'oid',
+ proargtypes => 'oid', prosrc => 'aggregate_dummy' },
{ oid => '2135', descr => 'minimum value of all float4 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13540', descr => 'minimum value of all float4 input values',
+ proname => 'min_p_float4', prokind => 'a', proisstrict => 'f', prorettype => 'float4',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2136', descr => 'minimum value of all float8 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13541', descr => 'minimum value of all float8 input values',
+ proname => 'min_p_float8', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2138', descr => 'minimum value of all date input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'date',
proargtypes => 'date', prosrc => 'aggregate_dummy' },
+{ oid => '13542', descr => 'minimum value of all date input values',
+ proname => 'min_p_date', prokind => 'a', proisstrict => 'f', prorettype => 'date',
+ proargtypes => 'date', prosrc => 'aggregate_dummy' },
{ oid => '2139', descr => 'minimum value of all time input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'time',
proargtypes => 'time', prosrc => 'aggregate_dummy' },
+{ oid => '13543', descr => 'minimum value of all time input values',
+ proname => 'min_p_time', prokind => 'a', proisstrict => 'f', prorettype => 'time',
+ proargtypes => 'time', prosrc => 'aggregate_dummy' },
{ oid => '2140',
descr => 'minimum value of all time with time zone input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
+{ oid => '13544',
+ descr => 'minimum value of all time with time zone input values',
+ proname => 'min_p_timetz', prokind => 'a', proisstrict => 'f', prorettype => 'timetz',
+ proargtypes => 'timetz', prosrc => 'aggregate_dummy' },
{ oid => '2141', descr => 'minimum value of all money input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'money',
proargtypes => 'money', prosrc => 'aggregate_dummy' },
+{ oid => '13545', descr => 'minimum value of all money input values',
+ proname => 'min_p_money', prokind => 'a', proisstrict => 'f', prorettype => 'money',
+ proargtypes => 'money', prosrc => 'aggregate_dummy' },
{ oid => '2142', descr => 'minimum value of all timestamp input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamp', proargtypes => 'timestamp',
prosrc => 'aggregate_dummy' },
+{ oid => '13546', descr => 'minimum value of all timestamp input values',
+ proname => 'min_p_timestamp', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamp', proargtypes => 'timestamp',
+ prosrc => 'aggregate_dummy' },
{ oid => '2143',
descr => 'minimum value of all timestamp with time zone input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'timestamptz', proargtypes => 'timestamptz',
prosrc => 'aggregate_dummy' },
+{ oid => '13547',
+ descr => 'minimum value of all timestamp with time zone input values',
+ proname => 'min_p_timestamptz', prokind => 'a', proisstrict => 'f',
+ prorettype => 'timestamptz', proargtypes => 'timestamptz',
+ prosrc => 'aggregate_dummy' },
{ oid => '2144', descr => 'minimum value of all interval input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13548', descr => 'minimum value of all interval input values',
+ proname => 'min_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => 'interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2145', descr => 'minimum value of all text values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'text',
proargtypes => 'text', prosrc => 'aggregate_dummy' },
+{ oid => '13549', descr => 'minimum value of all text values',
+ proname => 'min_p_text', prokind => 'a', proisstrict => 'f', prorettype => 'text',
+ proargtypes => 'text', prosrc => 'aggregate_dummy' },
{ oid => '2146', descr => 'minimum value of all numeric input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13550', descr => 'minimum value of all numeric input values',
+ proname => 'min_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2051', descr => 'minimum value of all anyarray input values',
proname => 'min', prokind => 'a', proisstrict => 'f',
prorettype => 'anyarray', proargtypes => 'anyarray',
prosrc => 'aggregate_dummy' },
+{ oid => '13551', descr => 'minimum value of all anyarray input values',
+ proname => 'min_p_anyarray', prokind => 'a', proisstrict => 'f',
+ prorettype => 'anyarray', proargtypes => 'anyarray',
+ prosrc => 'aggregate_dummy' },
{ oid => '2245', descr => 'minimum value of all bpchar input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
+{ oid => '13552', descr => 'minimum value of all bpchar input values',
+ proname => 'min_p_bpchar', prokind => 'a', proisstrict => 'f', prorettype => 'bpchar',
+ proargtypes => 'bpchar', prosrc => 'aggregate_dummy' },
{ oid => '2798', descr => 'minimum value of all tid input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
proargtypes => 'tid', prosrc => 'aggregate_dummy' },
+{ oid => '13553', descr => 'minimum value of all tid input values',
+ proname => 'min_p_tid', prokind => 'a', proisstrict => 'f', prorettype => 'tid',
+ proargtypes => 'tid', prosrc => 'aggregate_dummy' },
{ oid => '3565', descr => 'minimum value of all inet input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
proargtypes => 'inet', prosrc => 'aggregate_dummy' },
+{ oid => '13554', descr => 'minimum value of all inet input values',
+ proname => 'min_p_inet', prokind => 'a', proisstrict => 'f', prorettype => 'inet',
+ proargtypes => 'inet', prosrc => 'aggregate_dummy' },
{ oid => '4190', descr => 'minimum value of all pg_lsn input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
+{ oid => '13555', descr => 'minimum value of all pg_lsn input values',
+ proname => 'min_p_pg_lsn', prokind => 'a', proisstrict => 'f', prorettype => 'pg_lsn',
+ proargtypes => 'pg_lsn', prosrc => 'aggregate_dummy' },
{ oid => '5100', descr => 'minimum value of all xid8 input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
+{ oid => '13556', descr => 'minimum value of all xid8 input values',
+ proname => 'min_p_xid8', prokind => 'a', proisstrict => 'f', prorettype => 'xid8',
+ proargtypes => 'xid8', prosrc => 'aggregate_dummy' },
# count has two forms: count(any) and count(*)
{ oid => '2147',
@@ -6688,10 +6880,19 @@
proname => 'count', prosupport => 'int8inc_support', prokind => 'a',
proisstrict => 'f', prorettype => 'int8', proargtypes => 'any',
prosrc => 'aggregate_dummy' },
+{ oid => '13557',
+ descr => 'partial number of input rows for which the input expression is not null',
+ proname => 'count_p_any', prosupport => 'int8inc_support', prokind => 'a',
+ proisstrict => 'f', prorettype => 'int8', proargtypes => 'any',
+ prosrc => 'aggregate_dummy' },
{ oid => '2803', descr => 'number of input rows',
proname => 'count', prosupport => 'int8inc_support', prokind => 'a',
proisstrict => 'f', prorettype => 'int8', proargtypes => '',
prosrc => 'aggregate_dummy' },
+{ oid => '13558', descr => 'partial number of input rows',
+ proname => 'count_p', prosupport => 'int8inc_support', prokind => 'a',
+ proisstrict => 'f', prorettype => 'int8', proargtypes => '',
+ prosrc => 'aggregate_dummy' },
{ oid => '6236', descr => 'planner support for count run condition',
proname => 'int8inc_support', prorettype => 'internal',
proargtypes => 'internal', prosrc => 'int8inc_support' },
@@ -9081,9 +9282,15 @@
{ oid => '3526', descr => 'maximum value of all enum input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
+{ oid => '13559', descr => 'maximum value of all enum input values',
+ proname => 'max_p_anyenum', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
+ proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
{ oid => '3527', descr => 'minimum value of all enum input values',
proname => 'min', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
+{ oid => '13560', descr => 'minimum value of all enum input values',
+ proname => 'min_p_anyenum', prokind => 'a', proisstrict => 'f', prorettype => 'anyenum',
+ proargtypes => 'anyenum', prosrc => 'aggregate_dummy' },
{ oid => '3528', descr => 'first value of the input enum type',
proname => 'enum_first', proisstrict => 'f', provolatile => 's',
prorettype => 'anyenum', proargtypes => 'anyenum', prosrc => 'enum_first' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/expected/opr_sanity.out b/src/test/regress/expected/opr_sanity.out
index 330eb0f765..a3ddf2ff84 100644
--- a/src/test/regress/expected/opr_sanity.out
+++ b/src/test/regress/expected/opr_sanity.out
@@ -1722,14 +1722,58 @@ SELECT DISTINCT proname, oprname
FROM pg_operator AS o, pg_aggregate AS a, pg_proc AS p
WHERE a.aggfnoid = p.oid AND a.aggsortop = o.oid
ORDER BY 1, 2;
- proname | oprname
-----------+---------
- bool_and | <
- bool_or | >
- every | <
- max | >
- min | <
-(5 rows)
+ proname | oprname
+-------------------+---------
+ bool_and | <
+ bool_or | >
+ every | <
+ max | >
+ max_p_anyarray | >
+ max_p_anyenum | >
+ max_p_bpchar | >
+ max_p_date | >
+ max_p_floa8 | >
+ max_p_float4 | >
+ max_p_inet | >
+ max_p_int2 | >
+ max_p_int4 | >
+ max_p_int8 | >
+ max_p_interval | >
+ max_p_money | >
+ max_p_numeric | >
+ max_p_oid | >
+ max_p_pg_lsn | >
+ max_p_text | >
+ max_p_tid | >
+ max_p_time | >
+ max_p_timestamp | >
+ max_p_timestamptz | >
+ max_p_timetz | >
+ max_p_xid8 | >
+ min | <
+ min_p_anyarray | <
+ min_p_anyenum | <
+ min_p_bpchar | <
+ min_p_date | <
+ min_p_float4 | <
+ min_p_float8 | <
+ min_p_inet | <
+ min_p_int2 | <
+ min_p_int4 | <
+ min_p_int8 | <
+ min_p_interval | <
+ min_p_money | <
+ min_p_numeric | <
+ min_p_oid | <
+ min_p_pg_lsn | <
+ min_p_text | <
+ min_p_tid | <
+ min_p_time | <
+ min_p_timestamp | <
+ min_p_timestamptz | <
+ min_p_timetz | <
+ min_p_xid8 | <
+(49 rows)
-- Check datatypes match
SELECT a.aggfnoid::oid, o.oid
@@ -1762,14 +1806,58 @@ WHERE a.aggfnoid = p.oid AND a.aggsortop = o.oid AND
amopopr = o.oid AND
amopmethod = (SELECT oid FROM pg_am WHERE amname = 'btree')
ORDER BY 1, 2;
- proname | oprname | amopstrategy
-----------+---------+--------------
- bool_and | < | 1
- bool_or | > | 5
- every | < | 1
- max | > | 5
- min | < | 1
-(5 rows)
+ proname | oprname | amopstrategy
+-------------------+---------+--------------
+ bool_and | < | 1
+ bool_or | > | 5
+ every | < | 1
+ max | > | 5
+ max_p_anyarray | > | 5
+ max_p_anyenum | > | 5
+ max_p_bpchar | > | 5
+ max_p_date | > | 5
+ max_p_floa8 | > | 5
+ max_p_float4 | > | 5
+ max_p_inet | > | 5
+ max_p_int2 | > | 5
+ max_p_int4 | > | 5
+ max_p_int8 | > | 5
+ max_p_interval | > | 5
+ max_p_money | > | 5
+ max_p_numeric | > | 5
+ max_p_oid | > | 5
+ max_p_pg_lsn | > | 5
+ max_p_text | > | 5
+ max_p_tid | > | 5
+ max_p_time | > | 5
+ max_p_timestamp | > | 5
+ max_p_timestamptz | > | 5
+ max_p_timetz | > | 5
+ max_p_xid8 | > | 5
+ min | < | 1
+ min_p_anyarray | < | 1
+ min_p_anyenum | < | 1
+ min_p_bpchar | < | 1
+ min_p_date | < | 1
+ min_p_float4 | < | 1
+ min_p_float8 | < | 1
+ min_p_inet | < | 1
+ min_p_int2 | < | 1
+ min_p_int4 | < | 1
+ min_p_int8 | < | 1
+ min_p_interval | < | 1
+ min_p_money | < | 1
+ min_p_numeric | < | 1
+ min_p_oid | < | 1
+ min_p_pg_lsn | < | 1
+ min_p_text | < | 1
+ min_p_tid | < | 1
+ min_p_time | < | 1
+ min_p_timestamp | < | 1
+ min_p_timestamptz | < | 1
+ min_p_timetz | < | 1
+ min_p_xid8 | < | 1
+(49 rows)
-- Check that there are not aggregates with the same name and different
-- numbers of arguments. While not technically wrong, we have a project policy
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: [CAUTION!! freemail] Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 05:00 ` Re: Partial aggregates pushdown Ted Yu <[email protected]>
2022-11-22 09:11 ` RE: [CAUTION!! freemail] Re: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-22 10:51 ` Ted Yu <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Ted Yu @ 2022-11-22 10:51 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
On Tue, Nov 22, 2022 at 1:11 AM [email protected] <
[email protected]> wrote:
> Hi Mr.Yu.
>
> Thank you for comments.
>
> > + * Check that partial aggregate agg has compatibility
> >
> > If the `agg` refers to func parameter, the parameter name is aggform
> I fixed the above typo and made the above comment easy to understand
> New comment is "Check that partial aggregate function of aggform exsits in
> remote"
>
> > + int32 partialagg_minversion = PG_VERSION_NUM;
> > + if (aggform->partialagg_minversion ==
> > PARTIALAGG_MINVERSION_DEFAULT) {
> > + partialagg_minversion = PG_VERSION_NUM;
> >
> >
> > I am curious why the same variable is assigned the same value twice. It
> seems
> > the if block is redundant.
> >
> > + if ((fpinfo->server_version >= partialagg_minversion)) {
> > + compatible = true;
> >
> >
> > The above can be simplified as: return fpinfo->server_version >=
> > partialagg_minversion;
> I fixed according to your comment.
>
> Sincerely yours,
> Yuuki Fujii
>
>
> Hi,
Thanks for the quick response.
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-22 16:05 ` Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2022-11-22 16:05 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
[email protected] писал 2022-11-22 04:01:
> Hi Mr.Vondra, Mr.Pyhalov, Everyone.
>
> I discussed with Mr.Pyhalov about the above draft by directly sending
> mail to
> him(outside of pgsql-hackers). Mr.Pyhalov allowed me to update his
> patch
> along with the above draft. So I update Mr.Pyhalov's patch v10.
>
Hi, Yuki. Thank you for your work on this.
I've looked through the patch. Overall I like this approach, but have
the following comments.
1) Why should we require partialaggfn for min()/max()/count()? We could
just use original functions for a lot of aggregates, and so it would be
possible to push down some partial aggregates to older servers. I'm not
sure that it's a strict requirement, but a nice thing to think about.
Can we use the function itself as partialaggfn, for example, for
sum(int4)? For functions with internal aggtranstype (like sum(int8) it
would be more difficult).
2) fpinfo->server_version is not aggregated, for example, when we form
fpinfo in foreign_join_ok(), it seems we should spread it in more places
in postgres_fdw.c.
3) In add_foreign_grouping_paths() it seems there's no need for
additional argument, we can look at extra->patype. Also Assert() in
add_foreign_grouping_paths() will fire in --enable-cassert build.
4) Why do you modify lookup_agg_function() signature? I don't see tests,
showing that it's neccessary. Perhaps, more precise function naming
should be used instead?
5) In tests:
- Why version_num does have "name" type in
f_alter_server_version() function?
- You modify server_version option of 'loopback' server, but
don't reset it after test. This could affect further tests.
- "It's unsafe to push down partial aggregates with distinct"
in postgres_fdw.sql:3002 seems to be misleading.
3001
3002 -- It's unsafe to push down partial aggregates with distinct
3003 SELECT f_alter_server_version('loopback', 'set', -1);
3004 EXPLAIN (VERBOSE, COSTS OFF)
3005 SELECT avg(d) FROM pagg_tab;
3006 SELECT avg(d) FROM pagg_tab;
3007 select * from pg_foreign_server;
6) While looking at it, could cause a crash with something like
CREATE TYPE COMPLEX AS (re FLOAT, im FLOAT);
CREATE OR REPLACE FUNCTION
sum_complex (sum complex, el complex)
RETURNS complex AS
$$
DECLARE
s complex;
BEGIN
if el is not null and sum is not null then
sum.re:=coalesce(sum.re,0)+el.re;
sum.im:=coalesce(sum.im,0)+el.im;
end if;
RETURN sum;
END;
$$ LANGUAGE plpgSQL;
CREATE AGGREGATE SUM(COMPLEX) (
SFUNC=sum_complex,
STYPE=complex,
partialaggfunc=aaaa,
partialagg_minversion=1400
);
where aaaa - something nonexisting
enforce_generic_type_consistency (actual_arg_types=0x56269873d200,
declared_arg_types=0x0, nargs=1, rettype=0, allow_poly=true) at
parse_coerce.c:2132
2132 Oid decl_type =
declared_arg_types[j];
(gdb) bt
#0 enforce_generic_type_consistency (actual_arg_types=0x56269873d200,
declared_arg_types=0x0, nargs=1, rettype=0, allow_poly=true) at
parse_coerce.c:2132
#1 0x00005626960072de in lookup_agg_function (fnName=0x5626986715a0,
nargs=1, input_types=0x56269873d200, variadicArgType=0,
rettype=0x7ffd1a4045d8, only_normal=false) at pg_aggregate.c:916
#2 0x00005626960064ba in AggregateCreate (aggName=0x562698671000 "sum",
aggNamespace=2200, replace=false, aggKind=110 'n', numArgs=1,
numDirectArgs=0, parameterTypes=0x56269873d1e8, allParameterTypes=0,
parameterModes=0,
parameterNames=0, parameterDefaults=0x0, variadicArgType=0,
aggtransfnName=0x5626986712c0, aggfinalfnName=0x0, aggcombinefnName=0x0,
aggserialfnName=0x0, aggdeserialfnName=0x0, aggmtransfnName=0x0,
aggminvtransfnName=0x0,
aggmfinalfnName=0x0, partialaggfnName=0x5626986715a0,
finalfnExtraArgs=false, mfinalfnExtraArgs=false, finalfnModify=114 'r',
mfinalfnModify=114 'r', aggsortopName=0x0, aggTransType=16390,
aggTransSpace=0, aggmTransType=0,
aggmTransSpace=0, partialaggMinversion=1400, agginitval=0x0,
aggminitval=0x0, proparallel=117 'u') at pg_aggregate.c:582
#3 0x00005626960a1e1c in DefineAggregate (pstate=0x56269869ab48,
name=0x562698671038, args=0x5626986711b0, oldstyle=false,
parameters=0x5626986713b0, replace=false) at aggregatecmds.c:450
#4 0x000056269643061f in ProcessUtilitySlow (pstate=0x56269869ab48,
pstmt=0x562698671a68,
queryString=0x5626986705d8 "CREATE AGGREGATE SUM(COMPLEX)
(\nSFUNC=sum_complex,\nSTYPE=COMPLEX,\npartialaggfunc=scomplex,\npartialagg_minversion=1400\n);",
context=PROCESS_UTILITY_TOPLEVEL, params=0x0, queryEnv=0x0,
dest=0x562698671b48, qc=0x7ffd1a4053c0) at utility.c:1407
#5 0x000056269642fbb4 in standard_ProcessUtility (pstmt=0x562698671a68,
queryString=0x5626986705d8 "CREATE AGGREGATE SUM(COMPLEX)
(\nSFUNC=sum_complex,\nSTYPE=COMPLEX,\npartialaggfunc=scomplex,\npartialagg_minversion=1400\n);",
readOnlyTree=false, context=PROCESS_UTILITY_TOPLEVEL, params=0x0,
queryEnv=0x0, dest=0x562698671b48, qc=0x7ffd1a4053c0) at utility.c:1074
Later will look at it again.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2022-11-30 03:10 ` [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2022-11-30 03:10 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
Hi Mr.Pyhalov.
Thank you for comments.
> I've looked through the patch. Overall I like this approach, but have
> the following comments.
>
> 1) Why should we require partialaggfn for min()/max()/count()? We could
> just use original functions for a lot of aggregates, and so it would be
> possible to push down some partial aggregates to older servers. I'm not
> sure that it's a strict requirement, but a nice thing to think about.
> Can we use the function itself as partialaggfn, for example, for
> sum(int4)?
> For functions with internal aggtranstype (like sum(int8) it
> would be more difficult).
Thank you. I realized that partial aggregate pushdown is fine
without partialaggfn if original function has no aggfinalfn and
aggtranstype of it is not internal. So I have improved v12 by
this realization.
However, v13 requires partialaggfn for aggregate if it has aggfinalfn or
aggtranstype of it is internal such as sum(int8).
> 2) fpinfo->server_version is not aggregated, for example, when we form
> fpinfo in foreign_join_ok(), it seems we should spread it in more places
> in postgres_fdw.c.
I have responded to your comment by adding copy of server_version in
merge_fdw_options.
> 3) In add_foreign_grouping_paths() it seems there's no need for
> additional argument, we can look at extra->patype. Also Assert() in
> add_foreign_grouping_paths() will fire in --enable-cassert build.
I have fixed according to your comment.
> 4) Why do you modify lookup_agg_function() signature? I don't see tests,
> showing that it's neccessary. Perhaps, more precise function naming
> should be used instead?
I realized that there is no need of modification lookup_agg_function().
Instead, I use LookupFuncName().
> 5) In tests:
> - Why version_num does have "name" type in
> f_alter_server_version() function?
> - You modify server_version option of 'loopback' server, but
> don't reset it after test. This could affect further tests.
> - "It's unsafe to push down partial aggregates with distinct"
> in postgres_fdw.sql:3002 seems to be misleading.
> 3001
> 3002 -- It's unsafe to push down partial aggregates with distinct
> 3003 SELECT f_alter_server_version('loopback', 'set', -1);
I have fixed according to your comment.
> 6) While looking at it, could cause a crash with something like
I have fixed this problem by using LookupFuncName() instead of lookup_agg_function.
The following is readme of v13.
--readme of Partial aggregates push down v13
1. interface
1) pg_aggregate
There are the following additional columns.
a) partialaggfn
data type : regproc.
default value: zero(means invalid).
description : This field refers to the special aggregate function(then we call
this partialaggfunc)
corresponding to aggregation function(then we call src) which has aggfnoid.
partialaggfunc is used for partial aggregation pushdown by postgres_fdw.
The followings are differences between the src and the special aggregate function.
difference1) result type
The result type is same as the src's transtype if the src's transtype
is not internal.
Otherwise the result type is bytea.
difference2) final func
The final func does not exist if the src's transtype is not internal.
Otherwize the final func returns serialized value.
For example, there is a partialaggfunc avg_p_int4 which corresponds to avg(int4)
whose aggtranstype is _int4.
The result value of avg_p_int4 is a float8 array which consists of count and
summation. avg_p_int4 does not have finalfunc.
For another example, there is a partialaggfunc avg_p_int8 which corresponds to
avg(int8) whose aggtranstype is internal.
The result value of avg_p_int8 is a bytea serialized array which consists of count
and summation. avg_p_int8 has finalfunc int8_avg_serialize which is serialize function
of avg(int8). This field is zero if there is no partialaggfunc.
b) partialagg_minversion
data type : int4.
default value: zero(means current version).
description : This field is the minimum PostgreSQL server version which has
partialaggfunc. This field is used for checking compatibility of partialaggfunc.
The above fields are valid in tuples for builtin avg, sum, min, max, count.
There are additional records which correspond to partialaggfunc for avg, sum, min, max, count.
2) pg_proc
There are additional records which correspond to partialaggfunc for avg, sum, min, max, count.
3) postgres_fdw
postgres_fdw has an additional foreign server option server_version. server_version is
integer value which means remote server version number. Default value of server_version
is zero. server_version is used for checking compatibility of partialaggfunc.
2. feature
Partial aggregation pushdown is fine when either of the following conditions is true.
condition1) aggregate function has not internal aggtranstype and has no aggfinalfn.
condition2) the following two conditions are both true.
condition2-1) partialaggfn is valid.
condition2-2) server_version is not less than partialagg_minversion postgres_fdw executes
pushdown the patialaggfunc instead of a src.
postgres_fdw can pushdown partial aggregation of aggregate function which has internal
aggtranstype or has aggfinalfn if the function is one of avg, sum(int8), sum(numeric).
For example, we issue "select avg_p_int4(c) from t" instead of "select avg(c) from t"
in the above example.
postgres_fdw can pushdown every aggregate function which supports partial aggregation
if you add a partialaggfunc corresponding to the aggregate function by create aggregate command.
3. sample commands in psql
\c postgres
drop database tmp;
create database tmp;
\c tmp
create extension postgres_fdw;
create server server_01 foreign data wrapper postgres_fdw options(host 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
create user mapping for postgres server server_01 options(user 'postgres', password 'postgres');
create server server_02 foreign data wrapper postgres_fdw options(host 'localhost', dbname 'tmp', server_version '160000', async_capable 'true');
create user mapping for postgres server server_02 options(user 'postgres', password 'postgres');
create table t(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval) partition by list (type);
create table t1(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval);
create table t2(dt timestamp, id int4, name text, total int4, val float4, type int4, span interval);
truncate table t1;
truncate table t2;
insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 1, 1.1, 1, cast('1 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t1 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 2, 2.1, 1, cast('2 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 1, 1.1, 2, cast('1 seconds' as interval) from generate_series(1, 100000, 1) t;
insert into t2 select timestamp'2020-01-01' + cast(t || ' seconds' as interval), t % 100, 'hoge' || t, 2, 2.1, 2, cast('2 seconds' as interval) from generate_series(1, 100000, 1) t;
create foreign table f_t1 partition of t for values in (1) server server_01 options(table_name 't1');
create foreign table f_t2 partition of t for values in (2) server server_02 options(table_name 't2');
set enable_partitionwise_aggregate = on;
explain (verbose, costs off) select avg(total::int4), avg(total::int8) from t;
select avg(total::int4), avg(total::int8) from t;
--
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v13.patch (61.1K, ../../OS3PR01MB6660E2F999B97D8ADA32AE6B95159@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v13.patch)
download | inline diff:
From 591ce69344d60ca9c89f311b35d4d364ee7004ea Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Wed, 30 Nov 2022 09:25:01 +0900
Subject: [PATCH] Partial aggregates push down v13
---
contrib/postgres_fdw/deparse.c | 110 +++++-
.../postgres_fdw/expected/postgres_fdw.out | 318 +++++++++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 22 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 107 +++++-
src/backend/catalog/pg_aggregate.c | 32 ++
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 62 +++-
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 35 ++
src/test/regress/expected/oidjoins.out | 1 +
13 files changed, 724 insertions(+), 34 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..a404566a96 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,44 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", node->aggfnoid);
+ procform = (Form_pg_proc)GETSTRUCT(proctup);
+
+ if ((procform->prorettype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
+ appendFunctionName(node->aggfnoid, context);
+ } else if(aggform->partialaggfn) {
+ appendFunctionName((Oid)(aggform->partialaggfn), context);
+ } else {
+ elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(proctup);
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3992,68 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function of aggform exsits in remote
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ *
+ * condition1) agg is AGGKIND_NORMAL aggregate without
+ * distinct or order by clauses
+ * condition2) there is a aggregate function for partial aggregation
+ * (then we call this partialaggfunc) corresponding to agg.
+ * condition2 is true when either of the following cases is true.
+ * case2-1) return type of agg is not internal and agg has no finalfunc.
+ * In this case, partialaggfunc is agg itself.
+ * case2-2) agg has valid partialaggfn and partialagg_minversion <= server_version.
+ * In this case, partialaggfunc is a aggregate function whose oid is partialaggfn.
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ procform = (Form_pg_proc)GETSTRUCT(proctup);
+
+ if ((procform->prorettype == INTERNALOID) || (aggform->aggfinalfn != InvalidOid)) {
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn) {
+ partial_agg_ok = false;
+ } else if (!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+ }
+
+ ReleaseSysCache(aggtup);
+ ReleaseSysCache(proctup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..1109b528a9 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,285 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (sum((pagg_tab.d)::bigint)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | sum | avg | avg
+----+-----+-----+-------+------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | sum | avg | avg
+-----+-----+-------+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: postgres_fdw_sum((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ postgres_fdw_sum
+------------------
+ 58500
+(1 row)
+
+-- Error, invalid PARTIALAGGFUNC
+CREATE AGGREGATE postgres_fdw_sum_error(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_partialagg_foo,
+ PARTIALAGG_MINVERSION = 1500
+);
+ERROR: function postgres_fdw_partialagg_foo(bigint) does not exist
+-- Error, invalid PARTIALAGG_MINVERSION
+CREATE AGGREGATE postgres_fdw_sum_error(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_partialagg_foo,
+ PARTIALAGG_MINVERSION = aaa
+);
+ERROR: partialagg_minversion requires an integer value
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..7885e4eed8 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, PartitionwiseAggregateType patype);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -5980,6 +5983,7 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
fpinfo->async_capable = fpinfo_o->async_capable;
+ fpinfo->server_version = fpinfo_o->server_version;
/* Merge the table level options from either side of the join. */
if (fpinfo_i)
@@ -6159,7 +6163,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, PartitionwiseAggregateType patype)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6177,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6417,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6425,6 +6434,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6465,7 +6478,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6485,7 +6499,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..5bb35d9743 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for check compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..1c876652b9 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,87 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+
+-- Error, invalid PARTIALAGGFUNC
+CREATE AGGREGATE postgres_fdw_sum_error(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_partialagg_foo,
+ PARTIALAGG_MINVERSION = 1500
+);
+
+-- Error, invalid PARTIALAGG_MINVERSION
+CREATE AGGREGATE postgres_fdw_sum_error(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_partialagg_foo,
+ PARTIALAGG_MINVERSION = aaa
+);
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..fd533a5113 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -63,6 +63,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +73,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +93,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -569,6 +572,33 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = LookupFuncName(partialaggfnName, numArgs, aggArgTypes, false);
+
+ /* Check aggregate creator has permission to call the function */
+ aclresult = object_aclcheck(ProcedureRelationId, partialaggfn, GetUserId(), ACL_EXECUTE);
+ if (aclresult != ACLCHECK_OK)
+ aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(partialaggfn));
+
+ rettype = get_func_rettype(partialaggfn);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial aggregation function %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +714,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..19579bd3e2 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,27 +58,45 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
@@ -76,6 +116,11 @@
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
aggmtranstype => 'interval' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,7 +128,8 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index f15aa2dbb1..9af8ba4f9c 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6507,35 +6507,67 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
@@ -6558,6 +6590,9 @@
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-30 08:12 ` Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2022-11-30 08:12 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Zhihong Yu <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
Hi, Yuki.
1) In previous version of the patch aggregates, which had partialaggfn,
were ok to push down. And it was a definite sign that aggregate can be
pushed down. Now we allow pushing down an aggregate, which prorettype is
not internal and aggfinalfn is not defined. Is it safe for all
user-defined (or builtin) aggregates, even if they are generally
shippable? Aggcombinefn is executed locally and we check that aggregate
function itself is shippable. Is it enough? Perhaps, we could use
partialagg_minversion (like aggregates with partialagg_minversion == -1
should not be pushed down) or introduce separate explicit flag?
2) Do we really have to look at pg_proc in partial_agg_ok() and
deparseAggref()? Perhaps, looking at aggtranstype is enough?
3) I'm not sure if CREATE AGGREGATE tests with invalid
PARTIALAGGFUNC/PARTIALAGG_MINVERSION should be in postgres_fdw tests or
better should be moved to src/test/regress/sql/create_aggregate.sql, as
they are not specific to postgres_fdw
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2022-11-30 10:01 ` [email protected] <[email protected]>
2022-11-30 13:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 2 replies; 42+ messages in thread
From: [email protected] @ 2022-11-30 10:01 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Pyhalov.
> 1) In previous version of the patch aggregates, which had partialaggfn, were ok
> to push down. And it was a definite sign that aggregate can be pushed down.
> Now we allow pushing down an aggregate, which prorettype is not internal and
> aggfinalfn is not defined. Is it safe for all user-defined (or builtin) aggregates,
> even if they are generally shippable? Aggcombinefn is executed locally and we
> check that aggregate function itself is shippable. Is it enough? Perhaps, we
> could use partialagg_minversion (like aggregates with partialagg_minversion
> == -1 should not be pushed down) or introduce separate explicit flag?
In what case partial aggregate pushdown is unsafe for aggregate which has not internal aggtranstype
and has no aggfinalfn?
By reading [1], I believe that if aggcombinefn of such aggregate recieves return values of original
aggregate functions in each remote then it must produce same value that would have resulted
from scanning all the input in a single operation.
> 2) Do we really have to look at pg_proc in partial_agg_ok() and
> deparseAggref()? Perhaps, looking at aggtranstype is enough?
You are right. I fixed according to your comment.
> 3) I'm not sure if CREATE AGGREGATE tests with invalid
> PARTIALAGGFUNC/PARTIALAGG_MINVERSION should be in postgres_fdw
> tests or better should be moved to src/test/regress/sql/create_aggregate.sql,
> as they are not specific to postgres_fdw
Thank you. I moved these tests to src/test/regress/sql/create_aggregate.sql.
[1] https://www.postgresql.org/docs/15/xaggr.html#XAGGR-PARTIAL-AGGREGATES
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v14.patch (61.7K, ../../OS3PR01MB66604EFC363C61E6D589C53F95159@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v14.patch)
download | inline diff:
From 7bb630113eefc9cdf5ddfcf42204fc0c56feeb29 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Wed, 30 Nov 2022 17:58:23 +0900
Subject: [PATCH] Partial aggregates push down v14
---
contrib/postgres_fdw/deparse.c | 102 +++++-
.../postgres_fdw/expected/postgres_fdw.out | 294 +++++++++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 22 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 83 ++++-
src/backend/catalog/pg_aggregate.c | 32 ++
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 62 +++-
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 35 +++
.../regress/expected/create_aggregate.out | 24 ++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/sql/create_aggregate.sql | 24 ++
15 files changed, 716 insertions(+), 34 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..8e3341c657 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,36 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
+ appendFunctionName(node->aggfnoid, context);
+ } else if(aggform->partialaggfn) {
+ appendFunctionName((Oid)(aggform->partialaggfn), context);
+ } else {
+ elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3984,68 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function of aggform exsits in remote
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ *
+ * condition1) agg is AGGKIND_NORMAL aggregate without
+ * distinct or order by clauses
+ * condition2) there is a aggregate function for partial aggregation
+ * (then we call this partialaggfunc) corresponding to agg.
+ * condition2 is true when either of the following cases is true.
+ * case2-1) return type of agg is not internal and agg has no finalfunc.
+ * In this case, partialaggfunc is agg itself.
+ * case2-2) agg has valid partialaggfn and partialagg_minversion <= server_version.
+ * In this case, partialaggfunc is a aggregate function whose oid is partialaggfn.
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ procform = (Form_pg_proc)GETSTRUCT(proctup);
+
+ if ((procform->prorettype == INTERNALOID) || (aggform->aggfinalfn != InvalidOid)) {
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn) {
+ partial_agg_ok = false;
+ } else if (!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+ }
+
+ ReleaseSysCache(aggtup);
+ ReleaseSysCache(proctup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..effad116f5 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,261 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (sum((pagg_tab.d)::bigint)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | sum | avg | avg
+----+-----+-----+-------+------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | sum | avg | avg
+-----+-----+-------+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: postgres_fdw_sum((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ postgres_fdw_sum
+------------------
+ 58500
+(1 row)
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..7885e4eed8 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, PartitionwiseAggregateType patype);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -5980,6 +5983,7 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
fpinfo->async_capable = fpinfo_o->async_capable;
+ fpinfo->server_version = fpinfo_o->server_version;
/* Merge the table level options from either side of the join. */
if (fpinfo_i)
@@ -6159,7 +6163,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, PartitionwiseAggregateType patype)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6177,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6417,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6425,6 +6434,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6465,7 +6478,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6485,7 +6499,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..5bb35d9743 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for check compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..3bb187cf15 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,63 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..fd533a5113 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -63,6 +63,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +73,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +93,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -569,6 +572,33 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = LookupFuncName(partialaggfnName, numArgs, aggArgTypes, false);
+
+ /* Check aggregate creator has permission to call the function */
+ aclresult = object_aclcheck(ProcedureRelationId, partialaggfn, GetUserId(), ACL_EXECUTE);
+ if (aclresult != ACLCHECK_OK)
+ aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(partialaggfn));
+
+ rettype = get_func_rettype(partialaggfn);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial aggregation function %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +714,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..19579bd3e2 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,27 +58,45 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
@@ -76,6 +116,11 @@
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
aggmtranstype => 'interval' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,7 +128,8 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index f15aa2dbb1..9af8ba4f9c 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6507,35 +6507,67 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
@@ -6558,6 +6590,9 @@
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out
index dcf6909423..9e1a2393b1 100644
--- a/src/test/regress/expected/create_aggregate.out
+++ b/src/test/regress/expected/create_aggregate.out
@@ -322,3 +322,27 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized
WARNING: aggregate attribute "Finalfunc_modify" not recognized
WARNING: aggregate attribute "Parallel" not recognized
ERROR: aggregate stype must be specified
+-- invalid: PARTIALAGGFUNC which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 1600
+);
+ERROR: function partialagg_foo(bigint) does not exist
+-- invalid: PARTIALAGG_MINVERSION is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
+ERROR: partialagg_minversion requires an integer value
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql
index d4b4036fd7..7bc7d04e17 100644
--- a/src/test/regress/sql/create_aggregate.sql
+++ b/src/test/regress/sql/create_aggregate.sql
@@ -328,3 +328,27 @@ CREATE AGGREGATE case_agg(float8)
"Finalfunc_modify" = read_write,
"Parallel" = safe
);
+
+-- invalid: PARTIALAGGFUNC which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 1600
+);
+
+-- invalid: PARTIALAGG_MINVERSION is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
\ No newline at end of file
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-30 13:30 ` Alexander Pyhalov <[email protected]>
1 sibling, 0 replies; 42+ messages in thread
From: Alexander Pyhalov @ 2022-11-30 13:30 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
[email protected] писал 2022-11-30 13:01:
> Hi Mr.Pyhalov.
>
>> 1) In previous version of the patch aggregates, which had
>> partialaggfn, were ok
>> to push down. And it was a definite sign that aggregate can be pushed
>> down.
>> Now we allow pushing down an aggregate, which prorettype is not
>> internal and
>> aggfinalfn is not defined. Is it safe for all user-defined (or
>> builtin) aggregates,
>> even if they are generally shippable? Aggcombinefn is executed locally
>> and we
>> check that aggregate function itself is shippable. Is it enough?
>> Perhaps, we
>> could use partialagg_minversion (like aggregates with
>> partialagg_minversion
>> == -1 should not be pushed down) or introduce separate explicit flag?
> In what case partial aggregate pushdown is unsafe for aggregate which
> has not internal aggtranstype
> and has no aggfinalfn?
> By reading [1], I believe that if aggcombinefn of such aggregate
> recieves return values of original
> aggregate functions in each remote then it must produce same value
> that would have resulted
> from scanning all the input in a single operation.
>
One more issue I started to think about - now we don't check
partialagg_minversion for "simple" aggregates at all. Is it correct? It
seems that , for example, we could try to pushdown bit_or(int8) to old
servers, but it didn't exist, for example, in 8.4. I think it's a
broader issue (it would be also the case already if we push down
aggregates) and shouldn't be fixed here. But there is an issue -
is_shippable() is too optimistic.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-11-30 14:30 ` Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
1 sibling, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2022-11-30 14:30 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
[email protected] писал 2022-11-30 13:01:
>> 2) Do we really have to look at pg_proc in partial_agg_ok() and
>> deparseAggref()? Perhaps, looking at aggtranstype is enough?
> You are right. I fixed according to your comment.
>
partial_agg_ok() still looks at pg_proc.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2022-12-01 02:23 ` [email protected] <[email protected]>
2022-12-01 16:36 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2022-12-01 02:23 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Pyhalov.
> One more issue I started to think about - now we don't check
> partialagg_minversion for "simple" aggregates at all. Is it correct? It seems that ,
> for example, we could try to pushdown bit_or(int8) to old servers, but it didn't
> exist, for example, in 8.4. I think it's a broader issue (it would be also the case
> already if we push down
> aggregates) and shouldn't be fixed here. But there is an issue -
> is_shippable() is too optimistic.
I think it is correct for now.
F.38.7 of [1] says "A limitation however is that postgres_fdw generally assumes that
immutable built-in functions and operators are safe to send to the remote server for
execution, if they appear in a WHERE clause for a foreign table." and says that we can
avoid this limitation by rewriting query.
It looks that postgres_fdw follows this policy in case of UPPERREL_GROUP_AGG aggregate pushdown.
If a aggreagate has not internal aggtranstype and has not aggfinalfn ,
partialaggfn of it is equal to itself.
So I think that it is adequate to follow this policy in case of partial aggregate pushdown
for such aggregates.
> >> 2) Do we really have to look at pg_proc in partial_agg_ok() and
> >> deparseAggref()? Perhaps, looking at aggtranstype is enough?
> > You are right. I fixed according to your comment.
> >
>
> partial_agg_ok() still looks at pg_proc.
Sorry for taking up your time. I fixed it.
[1] https://www.postgresql.org/docs/current/postgres-fdw.html
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v15.patch (61.4K, ../../OS3PR01MB66603EFA3A42AB5EE74438E295149@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v15.patch)
download | inline diff:
From 7703117db5f2e74d0b9eceb651b15b29af5886c4 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Thu, 1 Dec 2022 10:02:11 +0900
Subject: [PATCH] Partial aggregates push down v15
---
contrib/postgres_fdw/deparse.c | 94 +++++-
.../postgres_fdw/expected/postgres_fdw.out | 294 +++++++++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 22 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 83 ++++-
src/backend/catalog/pg_aggregate.c | 32 ++
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 62 +++-
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 35 +++
.../regress/expected/create_aggregate.out | 24 ++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/sql/create_aggregate.sql | 24 ++
15 files changed, 708 insertions(+), 34 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..35f2d10237 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,36 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
+ appendFunctionName(node->aggfnoid, context);
+ } else if(aggform->partialaggfn) {
+ appendFunctionName((Oid)(aggform->partialaggfn), context);
+ } else {
+ elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3984,60 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function of aggform exsits in remote
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ *
+ * condition1) agg is AGGKIND_NORMAL aggregate without
+ * distinct or order by clauses
+ * condition2) there is a aggregate function for partial aggregation
+ * (then we call this partialaggfunc) corresponding to agg.
+ * condition2 is true when either of the following cases is true.
+ * case2-1) return type of agg is not internal and agg has no finalfunc.
+ * In this case, partialaggfunc is agg itself.
+ * case2-2) agg has valid partialaggfn and partialagg_minversion <= server_version.
+ * In this case, partialaggfunc is a aggregate function whose oid is partialaggfn.
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype == INTERNALOID) || (aggform->aggfinalfn != InvalidOid)) {
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn) {
+ partial_agg_ok = false;
+ } else if (!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..effad116f5 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,261 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (sum((pagg_tab.d)::bigint)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | sum | avg | avg
+----+-----+-----+-------+------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | sum | avg | avg
+-----+-----+-------+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: postgres_fdw_sum((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ postgres_fdw_sum
+------------------
+ 58500
+(1 row)
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..0dca06ccfe 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, PartitionwiseAggregateType patype);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -5980,6 +5983,7 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
fpinfo->async_capable = fpinfo_o->async_capable;
+ fpinfo->server_version = fpinfo_o->server_version;
/* Merge the table level options from either side of the join. */
if (fpinfo_i)
@@ -6159,7 +6163,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, PartitionwiseAggregateType patype)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6177,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6417,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6425,6 +6434,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6465,7 +6478,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6485,7 +6499,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..5bb35d9743 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for check compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..3bb187cf15 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,63 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates with distinct option
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates if remote server version is old
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+-- Partial aggregate for user defined aggregate is fine to push down
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 1500
+);
+
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..fd533a5113 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -63,6 +63,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +73,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +93,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -569,6 +572,33 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = LookupFuncName(partialaggfnName, numArgs, aggArgTypes, false);
+
+ /* Check aggregate creator has permission to call the function */
+ aclresult = object_aclcheck(ProcedureRelationId, partialaggfn, GetUserId(), ACL_EXECUTE);
+ if (aclresult != ACLCHECK_OK)
+ aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(partialaggfn));
+
+ rettype = get_func_rettype(partialaggfn);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial aggregation function %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +714,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..19579bd3e2 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,27 +58,45 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
@@ -76,6 +116,11 @@
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
aggmtranstype => 'interval' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,7 +128,8 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index f15aa2dbb1..9af8ba4f9c 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6507,35 +6507,67 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
@@ -6558,6 +6590,9 @@
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out
index dcf6909423..9e1a2393b1 100644
--- a/src/test/regress/expected/create_aggregate.out
+++ b/src/test/regress/expected/create_aggregate.out
@@ -322,3 +322,27 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized
WARNING: aggregate attribute "Finalfunc_modify" not recognized
WARNING: aggregate attribute "Parallel" not recognized
ERROR: aggregate stype must be specified
+-- invalid: PARTIALAGGFUNC which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 1600
+);
+ERROR: function partialagg_foo(bigint) does not exist
+-- invalid: PARTIALAGG_MINVERSION is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
+ERROR: partialagg_minversion requires an integer value
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql
index d4b4036fd7..7bc7d04e17 100644
--- a/src/test/regress/sql/create_aggregate.sql
+++ b/src/test/regress/sql/create_aggregate.sql
@@ -328,3 +328,27 @@ CREATE AGGREGATE case_agg(float8)
"Finalfunc_modify" = read_write,
"Parallel" = safe
);
+
+-- invalid: PARTIALAGGFUNC which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 1600
+);
+
+-- invalid: PARTIALAGG_MINVERSION is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
\ No newline at end of file
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-12-01 16:36 ` Alexander Pyhalov <[email protected]>
2022-12-05 02:03 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Alexander Pyhalov @ 2022-12-01 16:36 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
[email protected] писал 2022-12-01 05:23:
> Hi Mr.Pyhalov.
>
Hi.
Attaching minor fixes. I haven't proof-read all comments (but perhaps,
they need attention from some native speaker).
Tested it with queries from
https://github.com/swarm64/s64da-benchmark-toolkit, works as expected.
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 1.patch (1.1K, ../../[email protected]/2-1.patch)
download | inline diff:
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 35f2d102374..bd8a4acc112 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -3472,9 +3472,9 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
if ((aggform->aggtranstype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
appendFunctionName(node->aggfnoid, context);
} else if(aggform->partialaggfn) {
- appendFunctionName((Oid)(aggform->partialaggfn), context);
+ appendFunctionName(aggform->partialaggfn, context);
} else {
- elog(ERROR, "there in no partialaggfn %u", node->aggfnoid);
+ elog(ERROR, "there is no partialaggfn %u", node->aggfnoid);
}
ReleaseSysCache(aggtup);
}
@@ -3986,7 +3986,8 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
}
/*
- * Check that partial aggregate function of aggform exsits in remote
+ * Check that partial aggregate function, described by aggform,
+ * exists on remote server, described by fpinfo.
*/
static bool
partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-12-01 16:36 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
@ 2022-12-05 02:03 ` [email protected] <[email protected]>
2022-12-07 18:59 ` Re: Partial aggregates pushdown Andres Freund <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2022-12-05 02:03 UTC (permalink / raw)
To: Alexander Pyhalov <[email protected]>; +Cc: Tomas Vondra <[email protected]>; pgsql-hackers; Andres Freund <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Pyhalov.
> Attaching minor fixes. I haven't proof-read all comments (but perhaps, they
> need attention from some native speaker).
Thank you. I fixed according to your patch.
And I fixed have proof-read all comments and messages.
> Tested it with queries from
> https://github.com/swarm64/s64da-benchmark-toolkit, works as expected.
Thank you for additional tests.
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v16.patch (61.8K, ../../OS3PR01MB6660CB590AEF3644A9EEE07D95189@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v16.patch)
download | inline diff:
From cfcae152064a04fe81c708548387b81b570ca0b5 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Mon, 5 Dec 2022 10:31:23 +0900
Subject: [PATCH] Partial aggregates push down v16
---
contrib/postgres_fdw/deparse.c | 97 +++++-
.../postgres_fdw/expected/postgres_fdw.out | 296 +++++++++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 22 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 85 ++++-
src/backend/catalog/pg_aggregate.c | 32 ++
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 62 +++-
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 35 +++
.../regress/expected/create_aggregate.out | 24 ++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/sql/create_aggregate.sql | 24 ++
15 files changed, 715 insertions(+), 34 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..922027727f 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,37 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid or partialaggfn which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
+ appendFunctionName(node->aggfnoid, context);
+ } else if(aggform->partialaggfn) {
+ appendFunctionName(aggform->partialaggfn, context);
+ } else {
+ elog(ERROR, "there is no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3985,62 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function, described by aggform,
+ * exists on remote server, described by fpinfo.
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down.
+ *
+ * It is fine when all of the following conditions are true.
+ * condition1) agg is AGGKIND_NORMAL aggregate which contains no distinct
+ * or order by clauses
+ * condition2) there is an aggregate function for partial aggregation
+ * (then we call this partialaggfunc) corresponding to agg.
+ * condition2 is true when either of the following cases is true.
+ * case2-1) return type of agg is not internal and agg has no finalfunc.
+ * In this case, partialaggfunc is agg itself.
+ * case2-2) agg has valid partialaggfn and partialagg_minversion <= server_version.
+ * In this case, partialaggfunc is a aggregate function whose oid is partialaggfn.
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype == INTERNALOID) || (aggform->aggfinalfn != InvalidOid)) {
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn) {
+ partial_agg_ok = false;
+ } else if (!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..e31544606a 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,263 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (sum((pagg_tab.d)::bigint)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | sum | avg | avg
+----+-----+-----+-------+------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | sum | avg | avg
+-----+-----+-------+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates which contains distinct clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates when
+-- server_version is older than partialagg_minversion, described by pg_aggregate
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+-- It's safe to push down partial aggregate for user defined aggregate
+-- which has valid options.
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 150000
+);
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: postgres_fdw_sum((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ postgres_fdw_sum
+------------------
+ 58500
+(1 row)
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..0dca06ccfe 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, PartitionwiseAggregateType patype);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -5980,6 +5983,7 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
fpinfo->async_capable = fpinfo_o->async_capable;
+ fpinfo->server_version = fpinfo_o->server_version;
/* Merge the table level options from either side of the join. */
if (fpinfo_i)
@@ -6159,7 +6163,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, PartitionwiseAggregateType patype)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6177,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6417,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6425,6 +6434,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6465,7 +6478,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6485,7 +6499,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..01c4f45e32 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for checking compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..d6e2e50204 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,65 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates which contains distinct clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates when
+-- server_version is older than partialagg_minversion, described by pg_aggregate
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+-- It's safe to push down partial aggregate for user defined aggregate
+-- which has valid options.
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 150000
+);
+
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..39d53059e9 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -63,6 +63,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +73,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +93,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -569,6 +572,33 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = LookupFuncName(partialaggfnName, numArgs, aggArgTypes, false);
+
+ /* Check aggregate creator has permission to call the function */
+ aclresult = object_aclcheck(ProcedureRelationId, partialaggfn, GetUserId(), ACL_EXECUTE);
+ if (aclresult != ACLCHECK_OK)
+ aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(partialaggfn));
+
+ rettype = get_func_rettype(partialaggfn);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partialaggfunc of %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for partialaggfunc %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +714,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..19579bd3e2 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,27 +58,45 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
@@ -76,6 +116,11 @@
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
aggmtranstype => 'interval' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,7 +128,8 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index f15aa2dbb1..9af8ba4f9c 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6507,35 +6507,67 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
@@ -6558,6 +6590,9 @@
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out
index dcf6909423..3197c8abba 100644
--- a/src/test/regress/expected/create_aggregate.out
+++ b/src/test/regress/expected/create_aggregate.out
@@ -322,3 +322,27 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized
WARNING: aggregate attribute "Finalfunc_modify" not recognized
WARNING: aggregate attribute "Parallel" not recognized
ERROR: aggregate stype must be specified
+-- invalid: partialaggfunc which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 160000
+);
+ERROR: function partialagg_foo(bigint) does not exist
+-- invalid: partialagg_minversion is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
+ERROR: partialagg_minversion requires an integer value
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql
index d4b4036fd7..499c120ca5 100644
--- a/src/test/regress/sql/create_aggregate.sql
+++ b/src/test/regress/sql/create_aggregate.sql
@@ -328,3 +328,27 @@ CREATE AGGREGATE case_agg(float8)
"Finalfunc_modify" = read_write,
"Parallel" = safe
);
+
+-- invalid: partialaggfunc which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 160000
+);
+
+-- invalid: partialagg_minversion is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
\ No newline at end of file
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-12-01 16:36 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-05 02:03 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2022-12-07 18:59 ` Andres Freund <[email protected]>
2022-12-15 22:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Andres Freund @ 2022-12-07 18:59 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>
Hi,
On 2022-12-05 02:03:49 +0000, [email protected] wrote:
> > Attaching minor fixes. I haven't proof-read all comments (but perhaps, they
> > need attention from some native speaker).
> Thank you. I fixed according to your patch.
> And I fixed have proof-read all comments and messages.
cfbot complains about some compiler warnings when building with clang:
https://cirrus-ci.com/task/6606268580757504
deparse.c:3459:22: error: equality comparison with extraneous parentheses [-Werror,-Wparentheses-equality]
if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~
deparse.c:3459:22: note: remove extraneous parentheses around the comparison to silence this warning
if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
~ ^ ~
deparse.c:3459:22: note: use '=' to turn this equality comparison into an assignment
if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
^~
=
Greetings,
Andres Freund
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Re: Partial aggregates pushdown Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-01 21:31 ` Re: Partial aggregates pushdown Ilya Gladyshev <[email protected]>
2021-11-02 09:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Re: Partial aggregates pushdown Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 01:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-22 16:05 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 08:12 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-11-30 14:30 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-12-01 16:36 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2022-12-05 02:03 ` RE: Partial aggregates pushdown [email protected] <[email protected]>
2022-12-07 18:59 ` Re: Partial aggregates pushdown Andres Freund <[email protected]>
@ 2022-12-15 22:23 ` [email protected] <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: [email protected] @ 2022-12-15 22:23 UTC (permalink / raw)
To: Andres Freund <[email protected]>; +Cc: Alexander Pyhalov <[email protected]>; Tomas Vondra <[email protected]>; pgsql-hackers; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; Ilya Gladyshev <[email protected]>; [email protected] <[email protected]>
Hi Mr.Freund.
> cfbot complains about some compiler warnings when building with clang:
> https://cirrus-ci.com/task/6606268580757504
>
> deparse.c:3459:22: error: equality comparison with extraneous parentheses
> [-Werror,-Wparentheses-equality]
> if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
> ~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~
> deparse.c:3459:22: note: remove extraneous parentheses around the
> comparison to silence this warning
> if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
> ~ ^ ~
> deparse.c:3459:22: note: use '=' to turn this equality comparison into an
> assignment
> if ((node->aggsplit == AGGSPLIT_SIMPLE)) {
> ^~
> =
I fixed this error.
Sincerely yours,
Yuuki Fujii
--
Yuuki Fujii
Information Technology R&D Center Mitsubishi Electric Corporation
Attachments:
[application/octet-stream] 0001-Partial-aggregates-push-down-v17.patch (61.8K, ../../OS3PR01MB6660A1DAF6619B9A0BF7D39B95E19@OS3PR01MB6660.jpnprd01.prod.outlook.com/2-0001-Partial-aggregates-push-down-v17.patch)
download | inline diff:
From 71993e37093b3cec325de5989a03afb3073775aa Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Fri, 16 Dec 2022 09:33:30 +0900
Subject: [PATCH] Partial aggregates push down v17
---
contrib/postgres_fdw/deparse.c | 97 +++++-
.../postgres_fdw/expected/postgres_fdw.out | 296 +++++++++++++++++-
contrib/postgres_fdw/option.c | 24 ++
contrib/postgres_fdw/postgres_fdw.c | 22 +-
contrib/postgres_fdw/postgres_fdw.h | 3 +
contrib/postgres_fdw/sql/postgres_fdw.sql | 85 ++++-
src/backend/catalog/pg_aggregate.c | 32 ++
src/backend/commands/aggregatecmds.c | 8 +
src/bin/pg_dump/pg_dump.c | 25 +-
src/include/catalog/pg_aggregate.dat | 62 +++-
src/include/catalog/pg_aggregate.h | 11 +
src/include/catalog/pg_proc.dat | 35 +++
.../regress/expected/create_aggregate.out | 24 ++
src/test/regress/expected/oidjoins.out | 1 +
src/test/regress/sql/create_aggregate.sql | 24 ++
15 files changed, 715 insertions(+), 34 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 9524765650..8e5a45ceaa 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -202,7 +202,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref* agg, PgFdwRelationInfo* fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -907,8 +907,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3448,14 +3449,37 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
- /* Find aggregate name from aggfnoid which is a pg_proc entry */
- appendFunctionName(node->aggfnoid, context);
+ if (node->aggsplit == AGGSPLIT_SIMPLE) {
+ /* Find aggregate name from aggfnoid which is a pg_proc entry */
+ appendFunctionName(node->aggfnoid, context);
+ }
+ else {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ /* Find aggregate name from aggfnoid or partialaggfn which is a pg_proc entry */
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(node->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", node->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype != INTERNALOID) && (aggform->aggfinalfn == InvalidOid)) {
+ appendFunctionName(node->aggfnoid, context);
+ } else if(aggform->partialaggfn) {
+ appendFunctionName(aggform->partialaggfn, context);
+ } else {
+ elog(ERROR, "there is no partialaggfn %u", node->aggfnoid);
+ }
+ ReleaseSysCache(aggtup);
+ }
+
appendStringInfoChar(buf, '(');
/* Add DISTINCT */
@@ -3961,3 +3985,62 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate function, described by aggform,
+ * exists on remote server, described by fpinfo.
+ */
+static bool
+partial_agg_compatible(Form_pg_aggregate aggform, PgFdwRelationInfo *fpinfo)
+{
+ int32 partialagg_minversion = PG_VERSION_NUM;
+ if (aggform->partialagg_minversion != PARTIALAGG_MINVERSION_DEFAULT) {
+ partialagg_minversion = aggform->partialagg_minversion;
+ }
+ return (fpinfo->server_version >= partialagg_minversion);
+}
+
+/*
+ * Check that partial aggregate agg is fine to push down.
+ *
+ * It is fine when all of the following conditions are true.
+ * condition1) agg is AGGKIND_NORMAL aggregate which contains no distinct
+ * or order by clauses
+ * condition2) there is an aggregate function for partial aggregation
+ * (then we call this partialaggfunc) corresponding to agg.
+ * condition2 is true when either of the following cases is true.
+ * case2-1) return type of agg is not internal and agg has no finalfunc.
+ * In this case, partialaggfunc is agg itself.
+ * case2-2) agg has valid partialaggfn and partialagg_minversion <= server_version.
+ * In this case, partialaggfunc is a aggregate function whose oid is partialaggfn.
+ */
+static bool
+partial_agg_ok(Aggref* agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate)GETSTRUCT(aggtup);
+
+ if ((aggform->aggtranstype == INTERNALOID) || (aggform->aggfinalfn != InvalidOid)) {
+ /* Only aggregates which has partialaggfn, are allowed */
+ if (!aggform->partialaggfn) {
+ partial_agg_ok = false;
+ } else if (!partial_agg_compatible(aggform, fpinfo)) {
+ partial_agg_ok = false;
+ }
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 558e94b845..e31544606a 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9645,13 +9645,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+SELECT f_alter_server_version('loopback', 'add', 0);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9710,8 +9731,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9722,21 +9743,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9772,6 +9793,263 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (min(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), (sum(pagg_tab.d)), (sum((pagg_tab.d)::bigint)), (avg(pagg_tab.d)), (avg((pagg_tab.d)::bigint))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | min | max | count | sum | sum | avg | avg
+----+-----+-----+-------+------+------+---------------------+---------------------
+ 0 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 1 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 2 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 3 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 4 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 5 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 6 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 7 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 8 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 9 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 10 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 11 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 12 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 13 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 14 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 15 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 16 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 17 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 18 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 19 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 20 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 21 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 22 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 23 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 24 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 25 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 26 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 27 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 28 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 29 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 30 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 31 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 32 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 33 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 34 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 35 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 36 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 37 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 38 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 39 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+ 40 | 0 | 20 | 60 | 900 | 900 | 15.0000000000000000 | 15.0000000000000000
+ 41 | 1 | 21 | 60 | 960 | 960 | 16.0000000000000000 | 16.0000000000000000
+ 42 | 2 | 22 | 60 | 1020 | 1020 | 17.0000000000000000 | 17.0000000000000000
+ 43 | 3 | 23 | 60 | 1080 | 1080 | 18.0000000000000000 | 18.0000000000000000
+ 44 | 4 | 24 | 60 | 1140 | 1140 | 19.0000000000000000 | 19.0000000000000000
+ 45 | 5 | 25 | 60 | 1200 | 1200 | 20.0000000000000000 | 20.0000000000000000
+ 46 | 6 | 26 | 60 | 1260 | 1260 | 21.0000000000000000 | 21.0000000000000000
+ 47 | 7 | 27 | 60 | 1320 | 1320 | 22.0000000000000000 | 22.0000000000000000
+ 48 | 8 | 28 | 60 | 1380 | 1380 | 23.0000000000000000 | 23.0000000000000000
+ 49 | 9 | 29 | 60 | 1440 | 1440 | 24.0000000000000000 | 24.0000000000000000
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: min(pagg_tab.a), max(pagg_tab.a), count(*), sum(pagg_tab.d), sum((pagg_tab.d)::bigint), avg(pagg_tab.d), avg((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d)), (PARTIAL sum((pagg_tab.d)::bigint)), (PARTIAL avg(pagg_tab.d)), (PARTIAL avg((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d)), (PARTIAL sum((pagg_tab_1.d)::bigint)), (PARTIAL avg(pagg_tab_1.d)), (PARTIAL avg((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL min(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d)), (PARTIAL sum((pagg_tab_2.d)::bigint)), (PARTIAL avg(pagg_tab_2.d)), (PARTIAL avg((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT min(a), max(a), count(*), sum(d), sum_p_int8(d::bigint), avg_p_int4(d), avg_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+ min | max | count | sum | sum | avg | avg
+-----+-----+-------+-------+-------+---------------------+---------------------
+ 0 | 29 | 3000 | 58500 | 58500 | 19.5000000000000000 | 19.5000000000000000
+(1 row)
+
+-- It's unsafe to push down partial aggregates which contains distinct clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Merge Append
+ Sort Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It's unsafe to push down partial aggregates when
+-- server_version is older than partialagg_minversion, described by pg_aggregate
+SELECT f_alter_server_version('loopback', 'set', -1);
+ f_alter_server_version
+------------------------
+
+(1 row)
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+ QUERY PLAN
+------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(pagg_tab.d)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab.d)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p1
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_1.d)
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p2
+ -> Partial Aggregate
+ Output: PARTIAL avg(pagg_tab_2.d)
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.d
+ Remote SQL: SELECT d FROM public.pagg_tab_p3
+(18 rows)
+
+SELECT avg(d) FROM pagg_tab;
+ avg
+---------------------
+ 19.5000000000000000
+(1 row)
+
+-- It's safe to push down partial aggregate for user defined aggregate
+-- which has valid options.
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 150000
+);
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: postgres_fdw_sum((pagg_tab.d)::bigint)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_1.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL postgres_fdw_sum((pagg_tab_2.d)::bigint))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT public.postgres_fdw_sum_p_int8(d::bigint) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+ postgres_fdw_sum
+------------------
+ 58500
+(1 row)
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index fa80ee2a55..cb1b28baf2 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -210,6 +210,27 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
errmsg("sslcert and sslkey are superuser-only"),
errhint("User mappings with the sslcert or sslkey options set may only be created or modified by the superuser.")));
}
+ else if (strcmp(def->defname, "server_version") == 0)
+ {
+ char* value;
+ int int_val;
+ bool is_parsed;
+
+ value = defGetString(def);
+ is_parsed = parse_int(value, &int_val, 0, NULL);
+
+ if (!is_parsed)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid value for integer option \"%s\": %s",
+ def->defname, value)));
+
+ if (int_val <= 0)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("\"%s\" must be an integer value greater than zero",
+ def->defname)));
+ }
}
PG_RETURN_VOID();
@@ -265,6 +286,9 @@ InitPgFdwOptions(void)
{"sslcert", UserMappingRelationId, true},
{"sslkey", UserMappingRelationId, true},
+ /* options for partial aggregation pushdown */
+ {"server_version", ForeignServerRelationId, false},
+
{NULL, InvalidOid, false}
};
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 8d7500abfb..0dca06ccfe 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -518,7 +518,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, PartitionwiseAggregateType patype);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -651,6 +651,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
fpinfo->async_capable = false;
+ fpinfo->server_version = 0;
apply_server_options(fpinfo);
apply_table_options(fpinfo);
@@ -5922,6 +5923,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
(void) parse_int(defGetString(def), &fpinfo->fetch_size, 0, NULL);
else if (strcmp(def->defname, "async_capable") == 0)
fpinfo->async_capable = defGetBoolean(def);
+ else if (strcmp(def->defname, "server_version") == 0)
+ (void)parse_int(defGetString(def), &fpinfo->server_version, 0, NULL);
}
}
@@ -5980,6 +5983,7 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
fpinfo->async_capable = fpinfo_o->async_capable;
+ fpinfo->server_version = fpinfo_o->server_version;
/* Merge the table level options from either side of the join. */
if (fpinfo_i)
@@ -6159,7 +6163,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, PartitionwiseAggregateType patype)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6173,6 +6177,10 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+ /* It's unsafe to push having statements with partial aggregates */
+ if ((patype == PARTITIONWISE_AGGREGATE_PARTIAL) && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6409,6 +6417,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6425,6 +6434,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData*)extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6465,7 +6478,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6485,7 +6499,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, extra->patype))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index a11d45bedf..01c4f45e32 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -88,6 +88,9 @@ typedef struct PgFdwRelationInfo
int fetch_size; /* fetch size for this remote table */
+ /* Options for checking compatibility of partial aggregation */
+ int server_version;
+
/*
* Name of the relation, for use while EXPLAINing ForeignScan. It is used
* for join and upper relations but is set for all relations. For a base
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index b0dbb41fb5..d6e2e50204 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2919,16 +2919,34 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
+CREATE OR REPLACE FUNCTION f_server_version_lag(version_offset int4) RETURNS text AS $$
+ SELECT (setting::int4 + version_offset)::text FROM pg_settings WHERE name = 'server_version_num';
+$$ LANGUAGE sql;
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE OR REPLACE FUNCTION f_alter_server_version(server_name text, operation text, version_offset int4) RETURNS void AS $$
+ DECLARE
+ version_num text;
+ BEGIN
+ EXECUTE 'SELECT f_server_version_lag($1) '
+ INTO version_num
+ USING version_offset;
+ EXECUTE format('ALTER SERVER %I OPTIONS (%I server_version %L)',
+ server_name, operation, version_num);
+ RETURN;
+ END;
+$$ LANGUAGE plpgsql;
+
+SELECT f_alter_server_version('loopback', 'add', 0);
+
+CREATE TABLE pagg_tab (a int, b int, c text, d int4) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2962,6 +2980,65 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+SELECT min(a), max(a), count(*), sum(d), sum(d::int8), avg(d), avg(d::int8) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates which contains distinct clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It's unsafe to push down partial aggregates when
+-- server_version is older than partialagg_minversion, described by pg_aggregate
+SELECT f_alter_server_version('loopback', 'set', -1);
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(d) FROM pagg_tab;
+SELECT avg(d) FROM pagg_tab;
+
+-- It's safe to push down partial aggregate for user defined aggregate
+-- which has valid options.
+CREATE AGGREGATE postgres_fdw_sum_p_int8(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ FINALFUNC = int8_avg_serialize
+);
+
+CREATE AGGREGATE postgres_fdw_sum(int8) (
+ SFUNC = int8_avg_accum,
+ STYPE = internal,
+ COMBINEFUNC = int8_avg_combine,
+ FINALFUNC = numeric_poly_sum,
+ SERIALFUNC = int8_avg_serialize,
+ DESERIALFUNC = int8_avg_deserialize,
+ PARTIALAGGFUNC = postgres_fdw_sum_p_int8,
+ PARTIALAGG_MINVERSION = 150000
+);
+
+ALTER EXTENSION postgres_fdw ADD FUNCTION postgres_fdw_sum(int8);
+
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+SELECT postgres_fdw_sum(d::int8) FROM pagg_tab;
+
+ALTER EXTENSION postgres_fdw DROP FUNCTION postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum(int8);
+DROP AGGREGATE postgres_fdw_sum_p_int8(int8);
+
+DROP FUNCTION f_server_version_lag(int4);
+DROP FUNCTION f_alter_server_version(text, text, int4);
+ALTER SERVER loopback OPTIONS (DROP server_version);
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index a98445b741..39d53059e9 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -63,6 +63,7 @@ AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *partialaggfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -72,6 +73,7 @@ AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel)
@@ -91,6 +93,7 @@ AggregateCreate(const char *aggName,
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
+ Oid partialaggfn = InvalidOid; /* can be omitted */
Oid sortop = InvalidOid; /* can be omitted */
Oid *aggArgTypes = parameterTypes->values;
bool mtransIsStrict = false;
@@ -569,6 +572,33 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial aggregate function, if present.
+ */
+ if (partialaggfnName)
+ {
+ HeapTuple aggtup;
+ partialaggfn = LookupFuncName(partialaggfnName, numArgs, aggArgTypes, false);
+
+ /* Check aggregate creator has permission to call the function */
+ aclresult = object_aclcheck(ProcedureRelationId, partialaggfn, GetUserId(), ACL_EXECUTE);
+ if (aclresult != ACLCHECK_OK)
+ aclcheck_error(aclresult, OBJECT_FUNCTION, get_func_name(partialaggfn));
+
+ rettype = get_func_rettype(partialaggfn);
+
+ if (((aggTransType == INTERNALOID) && (rettype != BYTEAOID))
+ || ((aggTransType != INTERNALOID) && (rettype != aggTransType)))
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partialaggfunc of %s is not stype",
+ NameListToString(partialaggfnName))));
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(partialaggfn));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for partialaggfunc %u", partialaggfn);
+ ReleaseSysCache(aggtup);
+ }
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -684,6 +714,8 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggminitval - 1] = CStringGetTextDatum(aggminitval);
else
nulls[Anum_pg_aggregate_aggminitval - 1] = true;
+ values[Anum_pg_aggregate_partialaggfn - 1] = ObjectIdGetDatum(partialaggfn);
+ values[Anum_pg_aggregate_partialagg_minversion - 1] = Int32GetDatum(partialaggMinversion);
if (replace)
oldtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(procOid));
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index a9339e40b3..87b21ee1ee 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -70,6 +70,7 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialaggfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
@@ -83,6 +84,7 @@ DefineAggregate(ParseState *pstate,
TypeName *mtransType = NULL;
int32 transSpace = 0;
int32 mtransSpace = 0;
+ int32 partialaggMinversion = 0;
char *initval = NULL;
char *minitval = NULL;
char *parallel = NULL;
@@ -143,6 +145,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialaggfunc") == 0)
+ partialaggfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -182,6 +186,8 @@ DefineAggregate(ParseState *pstate,
mtransType = defGetTypeName(defel);
else if (strcmp(defel->defname, "msspace") == 0)
mtransSpace = defGetInt32(defel);
+ else if (strcmp(defel->defname, "partialagg_minversion") == 0)
+ partialaggMinversion = defGetInt32(defel);
else if (strcmp(defel->defname, "initcond") == 0)
initval = defGetString(defel);
else if (strcmp(defel->defname, "initcond1") == 0)
@@ -461,6 +467,7 @@ DefineAggregate(ParseState *pstate,
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
+ partialaggfuncName,
finalfuncExtraArgs,
mfinalfuncExtraArgs,
finalfuncModify,
@@ -470,6 +477,7 @@ DefineAggregate(ParseState *pstate,
transSpace, /* transition space */
mtransTypeId, /* transition data type */
mtransSpace, /* transition space */
+ partialaggMinversion,
initval, /* initial condition */
minitval, /* initial condition */
proparallel); /* parallel safe? */
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index da427f4d4a..2e2d38b25b 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -13406,6 +13406,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
+ const char *partialaggfn;
+ const char *partialagg_minversion;
bool aggfinalextra;
bool aggmfinalextra;
char aggfinalmodify;
@@ -13488,11 +13490,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 160000)
+ appendPQExpBufferStr(query,
+ "partialaggfn,\n"
+ "partialagg_minversion\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS partialaggfn,\n"
+ "0 AS partialagg_minversion\n");
appendPQExpBufferStr(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -13518,6 +13528,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ partialaggfn = PQgetvalue(res, 0, PQfnumber(res, "partialaggfn"));
+ partialagg_minversion = PQgetvalue(res, 0, PQfnumber(res, "partialagg_minversion"));
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -13607,6 +13619,15 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(partialaggfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALAGGFUNC = %s", partialaggfn);
+
+ if (strcmp(partialagg_minversion, "0") != 0)
+ {
+ appendPQExpBuffer(details, ",\n PARTIALAGG_MINVERSION = %s",
+ partialagg_minversion);
+ }
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index b9110a5298..19579bd3e2 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -13,22 +13,44 @@
[
# avg
+{ aggfnoid => 'avg_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal',
+ aggtransspace => '48' },
{ aggfnoid => 'avg(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_avg', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_avg', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'avg_p_int8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int4', aggtransfn => 'int4_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int4)', aggtransfn => 'int4_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_int2', aggtransfn => 'int2_avg_accum',
+ aggcombinefn => 'int4_avg_combine',
+ aggtranstype => '_int8',
+ agginitval => '{0,0}' },
{ aggfnoid => 'avg(int2)', aggtransfn => 'int2_avg_accum',
aggfinalfn => 'int8_avg', aggcombinefn => 'int4_avg_combine',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int8_avg', aggtranstype => '_int8', aggmtranstype => '_int8',
- agginitval => '{0,0}', aggminitval => '{0,0}' },
+ agginitval => '{0,0}', aggminitval => '{0,0}',
+ partialaggfn => 'avg_p_int2', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
{ aggfnoid => 'avg(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_avg', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -36,27 +58,45 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_avg', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'avg_p_numeric', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float4', aggtransfn => 'float4_accum',
+ aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float4)', aggtransfn => 'float4_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float4', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_float8', aggtransfn => 'float8_accum',
+ aggcombinefn => 'float8_combine',
aggtranstype => '_float8', agginitval => '{0,0,0}' },
{ aggfnoid => 'avg(float8)', aggtransfn => 'float8_accum',
aggfinalfn => 'float8_avg', aggcombinefn => 'float8_combine',
- aggtranstype => '_float8', agginitval => '{0,0,0}' },
+ aggtranstype => '_float8', agginitval => '{0,0,0}',
+ partialaggfn => 'avg_p_float8', partialagg_minversion => '160000' },
+{ aggfnoid => 'avg_p_interval', aggtransfn => 'interval_accum',
+ aggcombinefn => 'interval_combine',
+ aggtranstype => '_interval', agginitval => '{0 second,0 second}' },
{ aggfnoid => 'avg(interval)', aggtransfn => 'interval_accum',
aggfinalfn => 'interval_avg', aggcombinefn => 'interval_combine',
aggmtransfn => 'interval_accum', aggminvtransfn => 'interval_accum_inv',
aggmfinalfn => 'interval_avg', aggtranstype => '_interval',
aggmtranstype => '_interval', agginitval => '{0 second,0 second}',
- aggminitval => '{0 second,0 second}' },
+ aggminitval => '{0 second,0 second}',
+ partialaggfn => 'avg_p_interval', partialagg_minversion => '160000' },
# sum
+{ aggfnoid => 'sum_p_int8', aggtransfn => 'int8_avg_accum',
+ aggfinalfn => 'int8_avg_serialize', aggcombinefn => 'int8_avg_combine',
+ aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '48' },
{ aggfnoid => 'sum(int8)', aggtransfn => 'int8_avg_accum',
aggfinalfn => 'numeric_poly_sum', aggcombinefn => 'int8_avg_combine',
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ partialaggfn => 'sum_p_int8', partialagg_minversion => '160000' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
@@ -76,6 +116,11 @@
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
aggmtranstype => 'interval' },
+{ aggfnoid => 'sum_p_numeric', aggtransfn => 'numeric_avg_accum',
+ aggfinalfn => 'numeric_avg_serialize', aggcombinefn => 'numeric_avg_combine',
+ aggserialfn => 'numeric_avg_serialize',
+ aggdeserialfn => 'numeric_avg_deserialize',
+ aggtranstype => 'internal', aggtransspace => '128' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,7 +128,8 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128',
+ partialaggfn => 'sum_p_numeric', partialagg_minversion => '160000' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 593da9f76a..1b7bac0fa6 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,12 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* special aggregate function for partial aggregation (0 if none) */
+ regproc partialaggfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
+ /* minimum PostgreSQL's version which has compatible partialaggfn */
+ int32 partialagg_minversion BKI_DEFAULT(0);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -141,6 +147,9 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_aggregate_fnoid_index, 2650, AggregateFnoidIndexId,
#define AGGMODIFY_SHAREABLE 's'
#define AGGMODIFY_READ_WRITE 'w'
+/* Symbolic value for default partialagg_minversion */
+#define PARTIALAGG_MINVERSION_DEFAULT 0
+
#endif /* EXPOSE_TO_CLIENT_CODE */
@@ -164,6 +173,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
+ List *aggpartialfnName,
bool finalfnExtraArgs,
bool mfinalfnExtraArgs,
char finalfnModify,
@@ -173,6 +183,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggTransSpace,
Oid aggmTransType,
int32 aggmTransSpace,
+ int32 partialaggMinversion,
const char *agginitval,
const char *aggminitval,
char proparallel);
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index f15aa2dbb1..9af8ba4f9c 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -6507,35 +6507,67 @@
descr => 'the average (arithmetic mean) as numeric of all bigint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13500',
+ descr => 'the partial average (arithmetic mean) as numeric of all bigint values',
+ proname => 'avg_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2101',
descr => 'the average (arithmetic mean) as numeric of all integer values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
+{ oid => '13501',
+ descr => 'the partial average (arithmetic mean) as numeric of all integer values',
+ proname => 'avg_p_int4', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int4', prosrc => 'aggregate_dummy' },
{ oid => '2102',
descr => 'the average (arithmetic mean) as numeric of all smallint values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int2', prosrc => 'aggregate_dummy' },
+{ oid => '13502',
+ descr => 'the partial average (arithmetic mean) as numeric of all smallint values',
+ proname => 'avg_p_int2', prokind => 'a', proisstrict => 'f', prorettype => '_int8',
+ proargtypes => 'int2', prosrc => 'aggregate_dummy' },
{ oid => '2103',
descr => 'the average (arithmetic mean) as numeric of all numeric values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13503',
+ descr => 'the partial average (arithmetic mean) as numeric of all numeric values',
+ proname => 'avg_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2104',
descr => 'the average (arithmetic mean) as float8 of all float4 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float4', prosrc => 'aggregate_dummy' },
+{ oid => '13504',
+ descr => 'the partial average (arithmetic mean) as float8 of all float4 values',
+ proname => 'avg_p_float4', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float4', prosrc => 'aggregate_dummy' },
{ oid => '2105',
descr => 'the average (arithmetic mean) as float8 of all float8 values',
proname => 'avg', prokind => 'a', proisstrict => 'f', prorettype => 'float8',
proargtypes => 'float8', prosrc => 'aggregate_dummy' },
+{ oid => '13505',
+ descr => 'the partial average (arithmetic mean) as float8 of all float8 values',
+ proname => 'avg_p_float8', prokind => 'a', proisstrict => 'f', prorettype => '_float8',
+ proargtypes => 'float8', prosrc => 'aggregate_dummy' },
{ oid => '2106',
descr => 'the average (arithmetic mean) as interval of all interval values',
proname => 'avg', prokind => 'a', proisstrict => 'f',
prorettype => 'interval', proargtypes => 'interval',
prosrc => 'aggregate_dummy' },
+{ oid => '13506',
+ descr => 'the partial average (arithmetic mean) as interval of all interval values',
+ proname => 'avg_p_interval', prokind => 'a', proisstrict => 'f',
+ prorettype => '_interval', proargtypes => 'interval',
+ prosrc => 'aggregate_dummy' },
{ oid => '2107', descr => 'sum as numeric across all bigint input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'int8', prosrc => 'aggregate_dummy' },
+{ oid => '13507', descr => 'partial sum as numeric across all bigint input values',
+ proname => 'sum_p_int8', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'aggregate_dummy' },
{ oid => '2108', descr => 'sum as bigint across all integer input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
proargtypes => 'int4', prosrc => 'aggregate_dummy' },
@@ -6558,6 +6590,9 @@
{ oid => '2114', descr => 'sum as numeric across all numeric input values',
proname => 'sum', prokind => 'a', proisstrict => 'f', prorettype => 'numeric',
proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
+{ oid => '13508', descr => 'partial sum as numeric across all numeric input values',
+ proname => 'sum_p_numeric', prokind => 'a', proisstrict => 'f', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'aggregate_dummy' },
{ oid => '2115', descr => 'maximum value of all bigint input values',
proname => 'max', prokind => 'a', proisstrict => 'f', prorettype => 'int8',
diff --git a/src/test/regress/expected/create_aggregate.out b/src/test/regress/expected/create_aggregate.out
index dcf6909423..3197c8abba 100644
--- a/src/test/regress/expected/create_aggregate.out
+++ b/src/test/regress/expected/create_aggregate.out
@@ -322,3 +322,27 @@ WARNING: aggregate attribute "Finalfunc_extra" not recognized
WARNING: aggregate attribute "Finalfunc_modify" not recognized
WARNING: aggregate attribute "Parallel" not recognized
ERROR: aggregate stype must be specified
+-- invalid: partialaggfunc which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 160000
+);
+ERROR: function partialagg_foo(bigint) does not exist
+-- invalid: partialagg_minversion is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
+ERROR: partialagg_minversion requires an integer value
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 215eb899be..49f2c9bc35 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {partialaggfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
diff --git a/src/test/regress/sql/create_aggregate.sql b/src/test/regress/sql/create_aggregate.sql
index d4b4036fd7..499c120ca5 100644
--- a/src/test/regress/sql/create_aggregate.sql
+++ b/src/test/regress/sql/create_aggregate.sql
@@ -328,3 +328,27 @@ CREATE AGGREGATE case_agg(float8)
"Finalfunc_modify" = read_write,
"Parallel" = safe
);
+
+-- invalid: partialaggfunc which doesn't exist
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = 160000
+);
+
+-- invalid: partialagg_minversion is not integer
+CREATE AGGREGATE haspartialagg_agg(int8) (
+ sfunc = int8_avg_accum,
+ stype = internal,
+ combinefunc = int8_avg_combine,
+ finalfunc = numeric_poly_sum,
+ serialfunc = int8_avg_serialize,
+ deserialfunc = int8_avg_deserialize,
+ partialaggfunc = partialagg_foo,
+ partialagg_minversion = aaa
+);
\ No newline at end of file
--
2.30.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
@ 2021-10-21 21:43 Zhihong Yu <[email protected]>
2021-10-22 06:26 ` Re: Partial aggregates pushdown Alexander Pyhalov <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: Zhihong Yu @ 2021-10-21 21:43 UTC (permalink / raw)
To: PostgreSQL Developers <[email protected]>; Alexander Pyhalov <[email protected]>
Hi,
w.r.t. 0001-Partial-aggregates-push-down-v03.patch
For partial_agg_ok(),
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind !=
AGGKIND_NORMAL || agg->aggorder != NIL)
+ ok = false;
Since SearchSysCache1() is not called yet, you can return false directly.
+ if (aggform->aggpartialpushdownsafe != true)
The above can be written as:
if (!aggform->aggpartialpushdownsafe)
For build_conv_list():
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
...
+ }
+ convlist = lappend_oid(convlist, converter_oid);
Do you intend to append InvalidOid to convlist (when tlentry->expr is
not Aggref) ?
Cheers
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2021-10-21 21:43 Re: Partial aggregates pushdown Zhihong Yu <[email protected]>
@ 2021-10-22 06:26 ` Alexander Pyhalov <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Alexander Pyhalov @ 2021-10-22 06:26 UTC (permalink / raw)
To: Zhihong Yu <[email protected]>; +Cc: PostgreSQL Developers <[email protected]>
Zhihong Yu писал 2021-10-22 00:43:
> Hi,
> w.r.t. 0001-Partial-aggregates-push-down-v03.patch
>
Hi.
> For partial_agg_ok(),
>
> + if (agg->aggdistinct || agg->aggvariadic || agg->aggkind !=
> AGGKIND_NORMAL || agg->aggorder != NIL)
> + ok = false;
>
> Since SearchSysCache1() is not called yet, you can return false
> directly.
Fixed.
>
> + if (aggform->aggpartialpushdownsafe != true)
>
> The above can be written as:
>
> if (!aggform->aggpartialpushdownsafe)
Fixed.
>
> For build_conv_list():
>
> + Oid converter_oid = InvalidOid;
> +
> + if (IsA(tlentry->expr, Aggref))
>
> ...
> + }
> + convlist = lappend_oid(convlist, converter_oid);
>
> Do you intend to append InvalidOid to convlist (when tlentry->expr is
> not Aggref) ?
Yes, for each tlist member (which matches fpinfo->grouped_tlist in case
when foreignrel is UPPER_REL) we need to find corresponding converter.
If we don't append InvalidOid, we can't find convlist member,
corresponding to tlist member. Added comments to build_conv_list.
Also fixed error in pg_dump.c (we selected '0' when
aggpartialconverterfn was not defined in schema, but checked for '-').
--
Best regards,
Alexander Pyhalov,
Postgres Professional
Attachments:
[text/x-diff] 0001-Partial-aggregates-push-down-v04.patch (58.1K, ../../[email protected]/2-0001-Partial-aggregates-push-down-v04.patch)
download | inline diff:
From a18e2ff8de00592797e7c3ccb8d6cd536a2e4e72 Mon Sep 17 00:00:00 2001
From: Alexander Pyhalov <[email protected]>
Date: Thu, 14 Oct 2021 17:30:34 +0300
Subject: [PATCH] Partial aggregates push down
---
contrib/postgres_fdw/deparse.c | 46 +++-
.../postgres_fdw/expected/postgres_fdw.out | 185 +++++++++++++++-
contrib/postgres_fdw/postgres_fdw.c | 204 +++++++++++++++++-
contrib/postgres_fdw/sql/postgres_fdw.sql | 27 ++-
src/backend/catalog/pg_aggregate.c | 28 ++-
src/backend/commands/aggregatecmds.c | 23 +-
src/backend/utils/adt/numeric.c | 96 +++++++++
src/bin/pg_dump/pg_dump.c | 21 +-
src/include/catalog/catversion.h | 2 +-
src/include/catalog/pg_aggregate.dat | 106 ++++-----
src/include/catalog/pg_aggregate.h | 10 +-
src/include/catalog/pg_proc.dat | 6 +
src/test/regress/expected/oidjoins.out | 1 +
13 files changed, 672 insertions(+), 83 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index d98bd666818..fa9f487d66d 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -196,6 +196,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
+static bool partial_agg_ok(Aggref *agg);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -831,8 +832,10 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if ((agg->aggsplit != AGGSPLIT_SIMPLE) && (agg->aggsplit != AGGSPLIT_INITIAL_SERIAL))
+ return false;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg))
return false;
/* As usual, it must be shippable. */
@@ -3249,7 +3252,7 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
bool use_variadic;
/* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+ Assert(node->aggsplit == AGGSPLIT_SIMPLE || node->aggsplit == AGGSPLIT_INITIAL_SERIAL);
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
@@ -3719,3 +3722,40 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * Check that partial aggregate agg is fine to push down
+ */
+static bool
+partial_agg_ok(Aggref *agg)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggvariadic || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ /* Only aggregates, marked as pushdown safe, are allowed */
+ if (!aggform->aggpartialpushdownsafe)
+ ok = false;
+
+ /*
+ * But if an aggregate requires serialization/deserialization, partial
+ * converter should be defined
+ */
+ if (ok && agg->aggtranstype == INTERNALOID)
+ ok = (aggform->aggpartialconverterfn != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+
+ return ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index 44c4367b8f9..3e1b997875e 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -9279,13 +9279,13 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
@@ -9344,8 +9344,8 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
-------------------------------------------------------------------------
+ QUERY PLAN
+---------------------------------------------------------------------------
Sort
Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
@@ -9356,21 +9356,21 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
Filter: (avg(t1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p1 t1
Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p1
-> HashAggregate
Output: t1_1.a, count(((t1_1.*)::pagg_tab))
Group Key: t1_1.a
Filter: (avg(t1_1.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p2 t1_1
Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p2
-> HashAggregate
Output: t1_2.a, count(((t1_2.*)::pagg_tab))
Group Key: t1_2.a
Filter: (avg(t1_2.b) < '22'::numeric)
-> Foreign Scan on public.fpagg_tab_p3 t1_2
Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3
+ Remote SQL: SELECT a, b, c, d FROM public.pagg_tab_p3
(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
@@ -9406,6 +9406,173 @@ SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700
-> Foreign Scan on fpagg_tab_p3 pagg_tab_2
(15 rows)
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ QUERY PLAN
+------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a), PARTIAL count(*), PARTIAL sum(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a), PARTIAL count(*), PARTIAL sum(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a), PARTIAL count(*), PARTIAL sum(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(26 rows)
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a), count(*) FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max | count
+----+-----+-------
+ 0 | 20 | 60
+ 1 | 21 | 60
+ 2 | 22 | 60
+ 3 | 23 | 60
+ 4 | 24 | 60
+ 5 | 25 | 60
+ 6 | 26 | 60
+ 7 | 27 | 60
+ 8 | 28 | 60
+ 9 | 29 | 60
+ 10 | 20 | 60
+ 11 | 21 | 60
+ 12 | 22 | 60
+ 13 | 23 | 60
+ 14 | 24 | 60
+ 15 | 25 | 60
+ 16 | 26 | 60
+ 17 | 27 | 60
+ 18 | 28 | 60
+ 19 | 29 | 60
+ 20 | 20 | 60
+ 21 | 21 | 60
+ 22 | 22 | 60
+ 23 | 23 | 60
+ 24 | 24 | 60
+ 25 | 25 | 60
+ 26 | 26 | 60
+ 27 | 27 | 60
+ 28 | 28 | 60
+ 29 | 29 | 60
+ 30 | 20 | 60
+ 31 | 21 | 60
+ 32 | 22 | 60
+ 33 | 23 | 60
+ 34 | 24 | 60
+ 35 | 25 | 60
+ 36 | 26 | 60
+ 37 | 27 | 60
+ 38 | 28 | 60
+ 39 | 29 | 60
+ 40 | 20 | 60
+ 41 | 21 | 60
+ 42 | 22 | 60
+ 43 | 23 | 60
+ 44 | 24 | 60
+ 45 | 25 | 60
+ 46 | 26 | 60
+ 47 | 27 | 60
+ 48 | 28 | 60
+ 49 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(pagg_tab.a), count(*), sum(pagg_tab.d)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.d))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p1
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.d))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p2
+ -> Foreign Scan
+ Output: (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.d))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT max(a), count(*), sum(d) FROM public.pagg_tab_p3
+(15 rows)
+
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+ max | count | sum
+-----+-------+-------
+ 29 | 3000 | 58500
+(1 row)
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+---------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(12 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 45a09337d08..24411d987c1 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_class.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/vacuum.h"
@@ -48,6 +50,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -80,7 +83,12 @@ enum FdwScanPrivateIndex
* String describing join i.e. names of relations being joined and types
* of join, added when the scan is join
*/
- FdwScanPrivateRelations
+ FdwScanPrivateRelations,
+
+ /*
+ * List of functions to convert partial aggregate result
+ */
+ FdwScanPrivateConvertors
};
/*
@@ -139,10 +147,13 @@ typedef struct PgFdwScanState
* for a foreign join scan. */
TupleDesc tupdesc; /* tuple descriptor of scan */
AttInMetadata *attinmeta; /* attribute datatype conversion metadata */
+ AttInMetadata *rcvd_attinmeta; /* metadata for received tuples, NULL if
+ * there's no converters */
/* extracted fdw_private data */
char *query; /* text of SELECT command */
List *retrieved_attrs; /* list of retrieved attribute numbers */
+ List *conv_list; /* list of converters */
/* for remote query execution */
PGconn *conn; /* connection for the scan */
@@ -510,6 +521,7 @@ static HeapTuple make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context);
static void conversion_error_callback(void *arg);
@@ -517,7 +529,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ Node *havingQual, bool partial);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -526,7 +538,8 @@ static void add_paths_with_pathkeys_for_rel(PlannerInfo *root, RelOptInfo *rel,
static void add_foreign_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra);
+ GroupPathExtraData *extra,
+ bool partial);
static void add_foreign_ordered_paths(PlannerInfo *root,
RelOptInfo *input_rel,
RelOptInfo *ordered_rel);
@@ -541,6 +554,7 @@ static void merge_fdw_options(PgFdwRelationInfo *fpinfo,
const PgFdwRelationInfo *fpinfo_i);
static int get_batch_size_option(Relation rel);
+static List *build_conv_list(RelOptInfo *foreignrel);
/*
* Foreign-data wrapper handler function: return a struct with pointers
@@ -1233,6 +1247,7 @@ postgresGetForeignPlan(PlannerInfo *root,
List *local_exprs = NIL;
List *params_list = NIL;
List *fdw_scan_tlist = NIL;
+ List *fdw_conv_list = NIL;
List *fdw_recheck_quals = NIL;
List *retrieved_attrs;
StringInfoData sql;
@@ -1336,6 +1351,9 @@ postgresGetForeignPlan(PlannerInfo *root,
/* Build the list of columns to be fetched from the foreign server. */
fdw_scan_tlist = build_tlist_to_deparse(foreignrel);
+ /* Build the list of converters for partial aggregates. */
+ fdw_conv_list = build_conv_list(foreignrel);
+
/*
* Ensure that the outer plan produces a tuple whose descriptor
* matches our scan tuple slot. Also, remove the local conditions
@@ -1415,6 +1433,8 @@ postgresGetForeignPlan(PlannerInfo *root,
if (IS_JOIN_REL(foreignrel) || IS_UPPER_REL(foreignrel))
fdw_private = lappend(fdw_private,
makeString(fpinfo->relation_name));
+ if (IS_UPPER_REL(foreignrel))
+ fdw_private = lappend(fdw_private, fdw_conv_list);
/*
* Create the ForeignScan node for the given relation.
@@ -1433,6 +1453,48 @@ postgresGetForeignPlan(PlannerInfo *root,
outer_plan);
}
+/*
+ * Generate attinmeta if there are some converters:
+ * they are expecxted to return BYTEA, but real input type is likely different.
+ */
+static AttInMetadata *
+get_rcvd_attinmeta(TupleDesc tupdesc, List *conv_list)
+{
+ TupleDesc rcvd_tupdesc;
+
+ Assert(conv_list != NIL);
+
+ rcvd_tupdesc = CreateTupleDescCopy(tupdesc);
+ for (int i = 0; i < rcvd_tupdesc->natts; i++)
+ {
+ Oid converter = InvalidOid;
+ Form_pg_attribute att = TupleDescAttr(rcvd_tupdesc, i);
+
+ converter = list_nth_oid(conv_list, i);
+ if (converter != InvalidOid)
+ {
+ HeapTuple proctup;
+ Form_pg_proc procform;
+
+ proctup = SearchSysCache1(PROCOID, ObjectIdGetDatum(converter));
+
+ if (!HeapTupleIsValid(proctup))
+ elog(ERROR, "cache lookup failed for function %u", converter);
+
+ procform = (Form_pg_proc) GETSTRUCT(proctup);
+
+ if (procform->pronargs != 1)
+ elog(ERROR, "converter %s is expected to have one argument", NameStr(procform->proname));
+
+ att->atttypid = procform->proargtypes.values[0];
+
+ ReleaseSysCache(proctup);
+ }
+ }
+
+ return TupleDescGetAttInMetadata(rcvd_tupdesc);
+}
+
/*
* Construct a tuple descriptor for the scan tuples handled by a foreign join.
*/
@@ -1545,6 +1607,12 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
FdwScanPrivateSelectSql));
fsstate->retrieved_attrs = (List *) list_nth(fsplan->fdw_private,
FdwScanPrivateRetrievedAttrs);
+
+ if (list_length(fsplan->fdw_private) > FdwScanPrivateConvertors)
+ fsstate->conv_list = (List *) list_nth(fsplan->fdw_private,
+ FdwScanPrivateConvertors);
+
+
fsstate->fetch_size = intVal(list_nth(fsplan->fdw_private,
FdwScanPrivateFetchSize));
@@ -1572,6 +1640,8 @@ postgresBeginForeignScan(ForeignScanState *node, int eflags)
}
fsstate->attinmeta = TupleDescGetAttInMetadata(fsstate->tupdesc);
+ if (fsstate->conv_list != NIL)
+ fsstate->rcvd_attinmeta = get_rcvd_attinmeta(fsstate->tupdesc, fsstate->conv_list);
/*
* Prepare for processing of parameters used in remote query, if any.
@@ -3815,13 +3885,21 @@ fetch_more_data(ForeignScanState *node)
for (i = 0; i < numrows; i++)
{
+ AttInMetadata *attinmeta;
+
Assert(IsA(node->ss.ps.plan, ForeignScan));
+ if (fsstate->rcvd_attinmeta)
+ attinmeta = fsstate->rcvd_attinmeta;
+ else
+ attinmeta = fsstate->attinmeta;
+
fsstate->tuples[i] =
make_tuple_from_result_row(res, i,
fsstate->rel,
- fsstate->attinmeta,
+ attinmeta,
fsstate->retrieved_attrs,
+ fsstate->conv_list,
node,
fsstate->temp_cxt);
}
@@ -4309,6 +4387,7 @@ store_returning_result(PgFdwModifyState *fmstate,
fmstate->rel,
fmstate->attinmeta,
fmstate->retrieved_attrs,
+ NIL,
NULL,
fmstate->temp_cxt);
@@ -4603,6 +4682,7 @@ get_returning_data(ForeignScanState *node)
dmstate->rel,
dmstate->attinmeta,
dmstate->retrieved_attrs,
+ NIL,
node,
dmstate->temp_cxt);
ExecStoreHeapTuple(newtup, slot, false);
@@ -5183,6 +5263,7 @@ analyze_row_processor(PGresult *res, int row, PgFdwAnalyzeState *astate)
astate->rel,
astate->attinmeta,
astate->retrieved_attrs,
+ NIL,
NULL,
astate->temp_cxt);
@@ -6083,7 +6164,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ Node *havingQual, bool partial)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6097,6 +6178,11 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
if (query->groupingSets)
return false;
+
+ /* It's unsafe to push having statements with partial aggregates */
+ if (partial && havingQual)
+ return false;
+
/* Get the fpinfo of the underlying scan relation. */
ofpinfo = (PgFdwRelationInfo *) fpinfo->outerrel->fdw_private;
@@ -6336,6 +6422,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6350,7 +6437,11 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
{
case UPPERREL_GROUP_AGG:
add_foreign_grouping_paths(root, input_rel, output_rel,
- (GroupPathExtraData *) extra);
+ (GroupPathExtraData *) extra, false);
+ break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra, true);
break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
@@ -6375,7 +6466,8 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
static void
add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
RelOptInfo *grouped_rel,
- GroupPathExtraData *extra)
+ GroupPathExtraData *extra,
+ bool partial)
{
Query *parse = root->parse;
PgFdwRelationInfo *ifpinfo = input_rel->fdw_private;
@@ -6391,8 +6483,9 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
!root->hasHavingQual)
return;
- Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ Assert(((extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL) && !partial) ||
+ (extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL && partial));
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6412,7 +6505,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual, partial))
return;
/*
@@ -7108,6 +7201,30 @@ complete_pending_request(AsyncRequest *areq)
TupIsNull(areq->result) ? 0.0 : 1.0);
}
+/*
+ * Interface to fmgr to call converters
+ */
+static Datum
+call_converter(Oid converter, Oid collation, Datum value, bool isnull, bool *res_isnull)
+{
+ LOCAL_FCINFO(fcinfo, 1);
+ FmgrInfo flinfo;
+ Datum result;
+
+ fmgr_info(converter, &flinfo);
+
+ InitFunctionCallInfoData(*fcinfo, &flinfo, 1, collation, NULL, NULL);
+
+ fcinfo->args[0].value = value;
+ fcinfo->args[0].isnull = isnull;
+
+ result = FunctionCallInvoke(fcinfo);
+
+ if (res_isnull)
+ *res_isnull = fcinfo->isnull;
+ return result;
+}
+
/*
* Create a tuple from the specified row of the PGresult.
*
@@ -7127,6 +7244,7 @@ make_tuple_from_result_row(PGresult *res,
Relation rel,
AttInMetadata *attinmeta,
List *retrieved_attrs,
+ List *conv_list,
ForeignScanState *fsstate,
MemoryContext temp_context)
{
@@ -7178,6 +7296,7 @@ make_tuple_from_result_row(PGresult *res,
errcallback.previous = error_context_stack;
error_context_stack = &errcallback;
+
/*
* i indexes columns in the relation, j indexes columns in the PGresult.
*/
@@ -7209,6 +7328,20 @@ make_tuple_from_result_row(PGresult *res,
valstr,
attinmeta->attioparams[i - 1],
attinmeta->atttypmods[i - 1]);
+ if (conv_list != NIL)
+ {
+ Oid converter = list_nth_oid(conv_list, i - 1);
+
+ if (converter != InvalidOid)
+ {
+ Form_pg_attribute att = TupleDescAttr(tupdesc, i);
+ bool res_isnull;
+
+ values[i - 1] = call_converter(converter, att->attcollation, values[i - 1], nulls[i - 1], &res_isnull);
+
+ nulls[i - 1] = res_isnull;
+ }
+ }
}
else if (i == SelfItemPointerAttributeNumber)
{
@@ -7472,3 +7605,54 @@ get_batch_size_option(Relation rel)
return batch_size;
}
+
+/*
+ * For UPPER_REL build a list of converters, corresponding to tlist entries.
+ */
+static List *
+build_conv_list(RelOptInfo *foreignrel)
+{
+ List *conv_list = NIL;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) foreignrel->fdw_private;
+ ListCell *lc;
+
+ if (IS_UPPER_REL(foreignrel))
+ {
+ /* For UPPER_REL tlist matches grouped_tlist */
+ foreach(lc, fpinfo->grouped_tlist)
+ {
+ TargetEntry *tlentry = (TargetEntry *) lfirst(lc);
+ Oid converter_oid = InvalidOid;
+
+ if (IsA(tlentry->expr, Aggref))
+ {
+ Aggref *agg = (Aggref *) tlentry->expr;
+
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && agg->aggtranstype == INTERNALOID)
+ {
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for function %u", agg->aggfnoid);
+
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ converter_oid = aggform->aggpartialconverterfn;
+ Assert(converter_oid != InvalidOid);
+
+ ReleaseSysCache(aggtup);
+ }
+ }
+
+ /*
+ * We append InvalidOid to conv_list to preserve one-to-one
+ * mapping between tlist and conv_list members.
+ */
+ conv_list = lappend_oid(conv_list, converter_oid);
+ }
+ }
+
+ return conv_list;
+}
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index e7b869f8cea..db3f12f30be 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -2729,15 +2729,15 @@ RESET enable_partitionwise_join;
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+CREATE TABLE pagg_tab (a int, b int, c text, d numeric) PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i % 40 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -2771,6 +2771,25 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
EXPLAIN (COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+-- It's unsafe to push down having clause when there are partial aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are fine to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are fine to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+SELECT max(a), count(*), sum(d) FROM pagg_tab;
+
+-- Shouldn't try to push down
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/src/backend/catalog/pg_aggregate.c b/src/backend/catalog/pg_aggregate.c
index 1f63d8081b2..639ea4cf9a6 100644
--- a/src/backend/catalog/pg_aggregate.c
+++ b/src/backend/catalog/pg_aggregate.c
@@ -60,6 +60,7 @@ AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -74,7 +75,8 @@ AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel)
+ char proparallel,
+ bool partialPushdownSafe)
{
Relation aggdesc;
HeapTuple tup;
@@ -88,6 +90,7 @@ AggregateCreate(const char *aggName,
Oid combinefn = InvalidOid; /* can be omitted */
Oid serialfn = InvalidOid; /* can be omitted */
Oid deserialfn = InvalidOid; /* can be omitted */
+ Oid partialconverterfn = InvalidOid; /* can be omitted */
Oid mtransfn = InvalidOid; /* can be omitted */
Oid minvtransfn = InvalidOid; /* can be omitted */
Oid mfinalfn = InvalidOid; /* can be omitted */
@@ -569,6 +572,27 @@ AggregateCreate(const char *aggName,
format_type_be(finaltype))));
}
+ /*
+ * Validate the partial converter, if present.
+ */
+ if (aggpartialconverterfnName)
+ {
+ fnArgs[0] = finaltype;
+
+ partialconverterfn = lookup_agg_function(aggpartialconverterfnName, 1,
+ fnArgs, InvalidOid,
+ &rettype);
+
+ if (rettype != BYTEAOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_DATATYPE_MISMATCH),
+ errmsg("return type of partial serialization function %s is not %s",
+ NameListToString(aggserialfnName),
+ format_type_be(BYTEAOID))));
+
+ }
+
+
/* handle sortop, if supplied */
if (aggsortopName)
{
@@ -664,6 +688,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggcombinefn - 1] = ObjectIdGetDatum(combinefn);
values[Anum_pg_aggregate_aggserialfn - 1] = ObjectIdGetDatum(serialfn);
values[Anum_pg_aggregate_aggdeserialfn - 1] = ObjectIdGetDatum(deserialfn);
+ values[Anum_pg_aggregate_aggpartialconverterfn - 1] = ObjectIdGetDatum(partialconverterfn);
values[Anum_pg_aggregate_aggmtransfn - 1] = ObjectIdGetDatum(mtransfn);
values[Anum_pg_aggregate_aggminvtransfn - 1] = ObjectIdGetDatum(minvtransfn);
values[Anum_pg_aggregate_aggmfinalfn - 1] = ObjectIdGetDatum(mfinalfn);
@@ -676,6 +701,7 @@ AggregateCreate(const char *aggName,
values[Anum_pg_aggregate_aggtransspace - 1] = Int32GetDatum(aggTransSpace);
values[Anum_pg_aggregate_aggmtranstype - 1] = ObjectIdGetDatum(aggmTransType);
values[Anum_pg_aggregate_aggmtransspace - 1] = Int32GetDatum(aggmTransSpace);
+ values[Anum_pg_aggregate_aggpartialpushdownsafe - 1] = BoolGetDatum(partialPushdownSafe);
if (agginitval)
values[Anum_pg_aggregate_agginitval - 1] = CStringGetTextDatum(agginitval);
else
diff --git a/src/backend/commands/aggregatecmds.c b/src/backend/commands/aggregatecmds.c
index 046cf2df08f..81cb17a119f 100644
--- a/src/backend/commands/aggregatecmds.c
+++ b/src/backend/commands/aggregatecmds.c
@@ -69,11 +69,13 @@ DefineAggregate(ParseState *pstate,
List *combinefuncName = NIL;
List *serialfuncName = NIL;
List *deserialfuncName = NIL;
+ List *partialconverterfuncName = NIL;
List *mtransfuncName = NIL;
List *minvtransfuncName = NIL;
List *mfinalfuncName = NIL;
bool finalfuncExtraArgs = false;
bool mfinalfuncExtraArgs = false;
+ bool partialPushdownSafe = false;
char finalfuncModify = 0;
char mfinalfuncModify = 0;
List *sortoperatorName = NIL;
@@ -142,6 +144,8 @@ DefineAggregate(ParseState *pstate,
serialfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "deserialfunc") == 0)
deserialfuncName = defGetQualifiedName(defel);
+ else if (strcmp(defel->defname, "partialconverterfunc") == 0)
+ partialconverterfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "msfunc") == 0)
mtransfuncName = defGetQualifiedName(defel);
else if (strcmp(defel->defname, "minvfunc") == 0)
@@ -189,6 +193,8 @@ DefineAggregate(ParseState *pstate,
minitval = defGetString(defel);
else if (strcmp(defel->defname, "parallel") == 0)
parallel = defGetString(defel);
+ else if (strcmp(defel->defname, "partial_pushdown_safe") == 0)
+ partialPushdownSafe = defGetBoolean(defel);
else
ereport(WARNING,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -372,6 +378,18 @@ DefineAggregate(ParseState *pstate,
errmsg("must specify both or neither of serialization and deserialization functions")));
}
+ if (partialconverterfuncName)
+ {
+ /*
+ * Serialization is only needed/allowed for transtype INTERNAL.
+ */
+ if (transTypeId != INTERNALOID)
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
+ errmsg("serialization functions may be specified only when the aggregate transition data type is %s",
+ format_type_be(INTERNALOID))));
+ }
+
/*
* If a moving-aggregate transtype is specified, look that up. Same
* restrictions as for transtype.
@@ -457,6 +475,8 @@ DefineAggregate(ParseState *pstate,
combinefuncName, /* combine function name */
serialfuncName, /* serial function name */
deserialfuncName, /* deserial function name */
+ partialconverterfuncName, /* partial converter function
+ * name */
mtransfuncName, /* fwd trans function name */
minvtransfuncName, /* inv trans function name */
mfinalfuncName, /* final function name */
@@ -471,7 +491,8 @@ DefineAggregate(ParseState *pstate,
mtransSpace, /* transition space */
initval, /* initial condition */
minitval, /* initial condition */
- proparallel); /* parallel safe? */
+ proparallel, /* parallel safe? */
+ partialPushdownSafe); /* partial pushdown safe? */
}
/*
diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 1de744855f3..e26e3e2de41 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -5770,6 +5770,52 @@ int8_avg_deserialize(PG_FUNCTION_ARGS)
PG_RETURN_POINTER(result);
}
+/*
+ * int8_sum_to_internal_serialize
+ * Convert int8 argument to serialized internal representation
+ */
+Datum
+int8_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ PolyNumAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = (Int128AggState *) palloc0(sizeof(Int128AggState));
+ state->calcSumX2 = false;
+
+ if (!PG_ARGISNULL(0))
+ {
+#ifdef HAVE_INT128
+ do_int128_accum(state, (int128) PG_GETARG_INT64(0));
+#else
+ do_numeric_accum(state, int64_to_numeric(PG_GETARG_INT64(0)));
+#endif
+ }
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+#ifdef HAVE_INT128
+ int128_to_numericvar(state->sumX, &tmp_var);
+#else
+ accum_sum_final(&state->sumX, &tmp_var);
+#endif
+ numericvar_serialize(&buf, &tmp_var);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Inverse transition functions to go with the above.
*/
@@ -5995,6 +6041,56 @@ numeric_sum(PG_FUNCTION_ARGS)
PG_RETURN_NUMERIC(result);
}
+/*
+ * numeric_sum_to_internal_serialize
+ * Convert numeric argument to serialized internal representation
+ */
+Datum
+numeric_sum_to_internal_serialize(PG_FUNCTION_ARGS)
+{
+ NumericAggState *state = NULL;
+ StringInfoData buf;
+ bytea *result;
+ NumericVar tmp_var;
+
+ state = makeNumericAggStateCurrentContext(false);
+
+ if (!PG_ARGISNULL(0))
+ do_numeric_accum(state, PG_GETARG_NUMERIC(0));
+
+ init_var(&tmp_var);
+
+ pq_begintypsend(&buf);
+
+ /* N */
+ pq_sendint64(&buf, state->N);
+
+ /* sumX */
+ accum_sum_final(&state->sumX, &tmp_var);
+ numericvar_serialize(&buf, &tmp_var);
+
+ /* maxScale */
+ pq_sendint32(&buf, state->maxScale);
+
+ /* maxScaleCount */
+ pq_sendint64(&buf, state->maxScaleCount);
+
+ /* NaNcount */
+ pq_sendint64(&buf, state->NaNcount);
+
+ /* pInfcount */
+ pq_sendint64(&buf, state->pInfcount);
+
+ /* nInfcount */
+ pq_sendint64(&buf, state->nInfcount);
+
+ result = pq_endtypsend(&buf);
+
+ free_var(&tmp_var);
+
+ PG_RETURN_BYTEA_P(result);
+}
+
/*
* Workhorse routine for the standard deviance and variance
* aggregates. 'state' is aggregate's transition state.
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index a485fb2d070..18f1f894445 100644
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -14224,11 +14224,13 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
const char *aggcombinefn;
const char *aggserialfn;
const char *aggdeserialfn;
+ const char *aggpartialconverterfn;
const char *aggmtransfn;
const char *aggminvtransfn;
const char *aggmfinalfn;
bool aggfinalextra;
bool aggmfinalextra;
+ bool aggpartialpushdownsafe;
char aggfinalmodify;
char aggmfinalmodify;
const char *aggsortop;
@@ -14313,11 +14315,19 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (fout->remoteVersion >= 110000)
appendPQExpBufferStr(query,
"aggfinalmodify,\n"
- "aggmfinalmodify\n");
+ "aggmfinalmodify,\n");
else
appendPQExpBufferStr(query,
"'0' AS aggfinalmodify,\n"
- "'0' AS aggmfinalmodify\n");
+ "'0' AS aggmfinalmodify,\n");
+ if (fout->remoteVersion >= 150000)
+ appendPQExpBufferStr(query,
+ "aggpartialconverterfn,\n"
+ "aggpartialpushdownsafe\n");
+ else
+ appendPQExpBufferStr(query,
+ "'-' AS aggpartialconverterfn,\n"
+ "false as aggpartialpushdownsafe\n");
appendPQExpBuffer(query,
"FROM pg_catalog.pg_aggregate a, pg_catalog.pg_proc p "
@@ -14335,6 +14345,8 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
aggcombinefn = PQgetvalue(res, 0, PQfnumber(res, "aggcombinefn"));
aggserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggserialfn"));
aggdeserialfn = PQgetvalue(res, 0, PQfnumber(res, "aggdeserialfn"));
+ aggpartialconverterfn = PQgetvalue(res, 0, PQfnumber(res, "aggpartialconverterfn"));
+ aggpartialpushdownsafe = (PQgetvalue(res, 0, PQfnumber(res, "aggpartialpushdownsafe"))[0] == 't');
aggmtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggmtransfn"));
aggminvtransfn = PQgetvalue(res, 0, PQfnumber(res, "aggminvtransfn"));
aggmfinalfn = PQgetvalue(res, 0, PQfnumber(res, "aggmfinalfn"));
@@ -14429,6 +14441,11 @@ dumpAgg(Archive *fout, const AggInfo *agginfo)
if (strcmp(aggdeserialfn, "-") != 0)
appendPQExpBuffer(details, ",\n DESERIALFUNC = %s", aggdeserialfn);
+ if (strcmp(aggpartialconverterfn, "-") != 0)
+ appendPQExpBuffer(details, ",\n PARTIALCONVERTERFUNC = %s", aggpartialconverterfn);
+ if (aggpartialpushdownsafe)
+ appendPQExpBufferStr(details, ",\n PARTIAL_PUSHDOWN_SAFE");
+
if (strcmp(aggmtransfn, "-") != 0)
{
appendPQExpBuffer(details, ",\n MSFUNC = %s,\n MINVFUNC = %s,\n MSTYPE = %s",
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 3253b8751b1..187c81b4857 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
*/
/* yyyymmddN */
-#define CATALOG_VERSION_NO 202109101
+#define CATALOG_VERSION_NO 202110191
#endif
diff --git a/src/include/catalog/pg_aggregate.dat b/src/include/catalog/pg_aggregate.dat
index fc6d3bfd945..61c4d812b34 100644
--- a/src/include/catalog/pg_aggregate.dat
+++ b/src/include/catalog/pg_aggregate.dat
@@ -56,26 +56,27 @@
aggserialfn => 'int8_avg_serialize', aggdeserialfn => 'int8_avg_deserialize',
aggmtransfn => 'int8_avg_accum', aggminvtransfn => 'int8_avg_accum_inv',
aggmfinalfn => 'numeric_poly_sum', aggtranstype => 'internal',
- aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48' },
+ aggtransspace => '48', aggmtranstype => 'internal', aggmtransspace => '48',
+ aggpartialconverterfn => 'int8_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int4)', aggtransfn => 'int4_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int4_avg_accum', aggminvtransfn => 'int4_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(int2)', aggtransfn => 'int2_sum', aggcombinefn => 'int8pl',
aggmtransfn => 'int2_avg_accum', aggminvtransfn => 'int2_avg_accum_inv',
aggmfinalfn => 'int2int4_sum', aggtranstype => 'int8',
- aggmtranstype => '_int8', aggminitval => '{0,0}' },
+ aggmtranstype => '_int8', aggminitval => '{0,0}', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float4)', aggtransfn => 'float4pl',
- aggcombinefn => 'float4pl', aggtranstype => 'float4' },
+ aggcombinefn => 'float4pl', aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(float8)', aggtransfn => 'float8pl',
- aggcombinefn => 'float8pl', aggtranstype => 'float8' },
+ aggcombinefn => 'float8pl', aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(money)', aggtransfn => 'cash_pl', aggcombinefn => 'cash_pl',
aggmtransfn => 'cash_pl', aggminvtransfn => 'cash_mi',
- aggtranstype => 'money', aggmtranstype => 'money' },
+ aggtranstype => 'money', aggmtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(interval)', aggtransfn => 'interval_pl',
aggcombinefn => 'interval_pl', aggmtransfn => 'interval_pl',
aggminvtransfn => 'interval_mi', aggtranstype => 'interval',
- aggmtranstype => 'interval' },
+ aggmtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'sum(numeric)', aggtransfn => 'numeric_avg_accum',
aggfinalfn => 'numeric_sum', aggcombinefn => 'numeric_avg_combine',
aggserialfn => 'numeric_avg_serialize',
@@ -83,146 +84,149 @@
aggmtransfn => 'numeric_avg_accum', aggminvtransfn => 'numeric_accum_inv',
aggmfinalfn => 'numeric_sum', aggtranstype => 'internal',
aggtransspace => '128', aggmtranstype => 'internal',
- aggmtransspace => '128' },
+ aggmtransspace => '128', aggpartialconverterfn => 'numeric_sum_to_internal_serialize', aggpartialpushdownsafe => 't' },
# max
{ aggfnoid => 'max(int8)', aggtransfn => 'int8larger',
aggcombinefn => 'int8larger', aggsortop => '>(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int4)', aggtransfn => 'int4larger',
aggcombinefn => 'int4larger', aggsortop => '>(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(int2)', aggtransfn => 'int2larger',
aggcombinefn => 'int2larger', aggsortop => '>(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(oid)', aggtransfn => 'oidlarger',
aggcombinefn => 'oidlarger', aggsortop => '>(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float4)', aggtransfn => 'float4larger',
aggcombinefn => 'float4larger', aggsortop => '>(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(float8)', aggtransfn => 'float8larger',
aggcombinefn => 'float8larger', aggsortop => '>(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(date)', aggtransfn => 'date_larger',
aggcombinefn => 'date_larger', aggsortop => '>(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(time)', aggtransfn => 'time_larger',
aggcombinefn => 'time_larger', aggsortop => '>(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timetz)', aggtransfn => 'timetz_larger',
aggcombinefn => 'timetz_larger', aggsortop => '>(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(money)', aggtransfn => 'cashlarger',
aggcombinefn => 'cashlarger', aggsortop => '>(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamp)', aggtransfn => 'timestamp_larger',
aggcombinefn => 'timestamp_larger', aggsortop => '>(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(timestamptz)', aggtransfn => 'timestamptz_larger',
aggcombinefn => 'timestamptz_larger',
aggsortop => '>(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
{ aggfnoid => 'max(interval)', aggtransfn => 'interval_larger',
aggcombinefn => 'interval_larger', aggsortop => '>(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(text)', aggtransfn => 'text_larger',
aggcombinefn => 'text_larger', aggsortop => '>(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(numeric)', aggtransfn => 'numeric_larger',
aggcombinefn => 'numeric_larger', aggsortop => '>(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyarray)', aggtransfn => 'array_larger',
aggcombinefn => 'array_larger', aggsortop => '>(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(bpchar)', aggtransfn => 'bpchar_larger',
aggcombinefn => 'bpchar_larger', aggsortop => '>(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(tid)', aggtransfn => 'tidlarger',
aggcombinefn => 'tidlarger', aggsortop => '>(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(anyenum)', aggtransfn => 'enum_larger',
aggcombinefn => 'enum_larger', aggsortop => '>(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(inet)', aggtransfn => 'network_larger',
aggcombinefn => 'network_larger', aggsortop => '>(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'max(pg_lsn)', aggtransfn => 'pg_lsn_larger',
aggcombinefn => 'pg_lsn_larger', aggsortop => '>(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn', aggpartialpushdownsafe => 't' },
# min
{ aggfnoid => 'min(int8)', aggtransfn => 'int8smaller',
aggcombinefn => 'int8smaller', aggsortop => '<(int8,int8)',
- aggtranstype => 'int8' },
+ aggtranstype => 'int8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int4)', aggtransfn => 'int4smaller',
aggcombinefn => 'int4smaller', aggsortop => '<(int4,int4)',
- aggtranstype => 'int4' },
+ aggtranstype => 'int4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(int2)', aggtransfn => 'int2smaller',
aggcombinefn => 'int2smaller', aggsortop => '<(int2,int2)',
- aggtranstype => 'int2' },
+ aggtranstype => 'int2', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(oid)', aggtransfn => 'oidsmaller',
aggcombinefn => 'oidsmaller', aggsortop => '<(oid,oid)',
- aggtranstype => 'oid' },
+ aggtranstype => 'oid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float4)', aggtransfn => 'float4smaller',
aggcombinefn => 'float4smaller', aggsortop => '<(float4,float4)',
- aggtranstype => 'float4' },
+ aggtranstype => 'float4', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(float8)', aggtransfn => 'float8smaller',
aggcombinefn => 'float8smaller', aggsortop => '<(float8,float8)',
- aggtranstype => 'float8' },
+ aggtranstype => 'float8', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(date)', aggtransfn => 'date_smaller',
aggcombinefn => 'date_smaller', aggsortop => '<(date,date)',
- aggtranstype => 'date' },
+ aggtranstype => 'date', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(time)', aggtransfn => 'time_smaller',
aggcombinefn => 'time_smaller', aggsortop => '<(time,time)',
- aggtranstype => 'time' },
+ aggtranstype => 'time', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timetz)', aggtransfn => 'timetz_smaller',
aggcombinefn => 'timetz_smaller', aggsortop => '<(timetz,timetz)',
- aggtranstype => 'timetz' },
+ aggtranstype => 'timetz', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(money)', aggtransfn => 'cashsmaller',
aggcombinefn => 'cashsmaller', aggsortop => '<(money,money)',
- aggtranstype => 'money' },
+ aggtranstype => 'money', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamp)', aggtransfn => 'timestamp_smaller',
aggcombinefn => 'timestamp_smaller', aggsortop => '<(timestamp,timestamp)',
- aggtranstype => 'timestamp' },
+ aggtranstype => 'timestamp', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(timestamptz)', aggtransfn => 'timestamptz_smaller',
aggcombinefn => 'timestamptz_smaller',
- aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz' },
+ aggsortop => '<(timestamptz,timestamptz)', aggtranstype => 'timestamptz',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(interval)', aggtransfn => 'interval_smaller',
aggcombinefn => 'interval_smaller', aggsortop => '<(interval,interval)',
- aggtranstype => 'interval' },
+ aggtranstype => 'interval', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(text)', aggtransfn => 'text_smaller',
aggcombinefn => 'text_smaller', aggsortop => '<(text,text)',
- aggtranstype => 'text' },
+ aggtranstype => 'text', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(numeric)', aggtransfn => 'numeric_smaller',
aggcombinefn => 'numeric_smaller', aggsortop => '<(numeric,numeric)',
- aggtranstype => 'numeric' },
+ aggtranstype => 'numeric', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyarray)', aggtransfn => 'array_smaller',
aggcombinefn => 'array_smaller', aggsortop => '<(anyarray,anyarray)',
- aggtranstype => 'anyarray' },
+ aggtranstype => 'anyarray', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(bpchar)', aggtransfn => 'bpchar_smaller',
aggcombinefn => 'bpchar_smaller', aggsortop => '<(bpchar,bpchar)',
- aggtranstype => 'bpchar' },
+ aggtranstype => 'bpchar', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(tid)', aggtransfn => 'tidsmaller',
aggcombinefn => 'tidsmaller', aggsortop => '<(tid,tid)',
- aggtranstype => 'tid' },
+ aggtranstype => 'tid', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(anyenum)', aggtransfn => 'enum_smaller',
aggcombinefn => 'enum_smaller', aggsortop => '<(anyenum,anyenum)',
- aggtranstype => 'anyenum' },
+ aggtranstype => 'anyenum', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(inet)', aggtransfn => 'network_smaller',
aggcombinefn => 'network_smaller', aggsortop => '<(inet,inet)',
- aggtranstype => 'inet' },
+ aggtranstype => 'inet', aggpartialpushdownsafe => 't' },
{ aggfnoid => 'min(pg_lsn)', aggtransfn => 'pg_lsn_smaller',
aggcombinefn => 'pg_lsn_smaller', aggsortop => '<(pg_lsn,pg_lsn)',
- aggtranstype => 'pg_lsn' },
+ aggtranstype => 'pg_lsn' , aggpartialpushdownsafe => 't'},
# count
{ aggfnoid => 'count(any)', aggtransfn => 'int8inc_any',
aggcombinefn => 'int8pl', aggmtransfn => 'int8inc_any',
aggminvtransfn => 'int8dec_any', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
{ aggfnoid => 'count()', aggtransfn => 'int8inc', aggcombinefn => 'int8pl',
aggmtransfn => 'int8inc', aggminvtransfn => 'int8dec', aggtranstype => 'int8',
- aggmtranstype => 'int8', agginitval => '0', aggminitval => '0' },
+ aggmtranstype => 'int8', agginitval => '0', aggminitval => '0',
+ aggpartialpushdownsafe => 't' },
# var_pop
{ aggfnoid => 'var_pop(int8)', aggtransfn => 'int8_accum',
diff --git a/src/include/catalog/pg_aggregate.h b/src/include/catalog/pg_aggregate.h
index 08c9379ba77..2c63102ff31 100644
--- a/src/include/catalog/pg_aggregate.h
+++ b/src/include/catalog/pg_aggregate.h
@@ -55,6 +55,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* function to convert bytea to transtype (0 if none) */
regproc aggdeserialfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+ /* function to convert aggregate result to bytea (0 if none) */
+ regproc aggpartialconverterfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
+
/* forward function for moving-aggregate mode (0 if none) */
regproc aggmtransfn BKI_DEFAULT(-) BKI_LOOKUP_OPT(pg_proc);
@@ -91,6 +94,9 @@ CATALOG(pg_aggregate,2600,AggregateRelationId)
/* estimated size of moving-agg state (0 for default est) */
int32 aggmtransspace BKI_DEFAULT(0);
+ /* true if partial aggregate is fine to push down */
+ bool aggpartialpushdownsafe BKI_DEFAULT(f);
+
#ifdef CATALOG_VARLEN /* variable-length fields start here */
/* initial value for transition state (can be NULL) */
@@ -161,6 +167,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
List *aggcombinefnName,
List *aggserialfnName,
List *aggdeserialfnName,
+ List *aggpartialconverterfnName,
List *aggmtransfnName,
List *aggminvtransfnName,
List *aggmfinalfnName,
@@ -175,6 +182,7 @@ extern ObjectAddress AggregateCreate(const char *aggName,
int32 aggmTransSpace,
const char *agginitval,
const char *aggminitval,
- char proparallel);
+ char proparallel,
+ bool partialPushdownSafe);
#endif /* PG_AGGREGATE_H */
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
index d068d6532ec..1cf23b15df0 100644
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -4798,6 +4798,12 @@
{ oid => '2786', descr => 'aggregate serial function',
proname => 'int8_avg_serialize', prorettype => 'bytea',
proargtypes => 'internal', prosrc => 'int8_avg_serialize' },
+{ oid => '9630', descr => 'partial aggregate converter function',
+ proname => 'int8_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'int8', prosrc => 'int8_sum_to_internal_serialize' },
+{ oid => '9631', descr => 'partial aggregate converter function',
+ proname => 'numeric_sum_to_internal_serialize', prorettype => 'bytea',
+ proargtypes => 'numeric', prosrc => 'numeric_sum_to_internal_serialize' },
{ oid => '2787', descr => 'aggregate deserial function',
proname => 'int8_avg_deserialize', prorettype => 'internal',
proargtypes => 'bytea internal', prosrc => 'int8_avg_deserialize' },
diff --git a/src/test/regress/expected/oidjoins.out b/src/test/regress/expected/oidjoins.out
index 1461e947cdf..e1680054c30 100644
--- a/src/test/regress/expected/oidjoins.out
+++ b/src/test/regress/expected/oidjoins.out
@@ -145,6 +145,7 @@ NOTICE: checking pg_aggregate {aggfinalfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggcombinefn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggserialfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggdeserialfn} => pg_proc {oid}
+NOTICE: checking pg_aggregate {aggpartialconverterfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggminvtransfn} => pg_proc {oid}
NOTICE: checking pg_aggregate {aggmfinalfn} => pg_proc {oid}
--
2.25.1
^ permalink raw reply [nested|flat] 42+ messages in thread
* [PATCH] add ruleutils.c support for MERGE
@ 2023-05-05 18:40 Alvaro Herrera <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Alvaro Herrera @ 2023-05-05 18:40 UTC (permalink / raw)
---
src/backend/utils/adt/ruleutils.c | 119 ++++++++++++++++++++++++++++
src/test/regress/expected/merge.out | 68 ++++++++++++++++
src/test/regress/sql/merge.sql | 38 +++++++++
3 files changed, 225 insertions(+)
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
index 461735e84f0..851488dd6d3 100644
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -411,6 +411,8 @@ static void get_update_query_targetlist_def(Query *query, List *targetList,
RangeTblEntry *rte);
static void get_delete_query_def(Query *query, deparse_context *context,
bool colNamesVisible);
+static void get_merge_query_def(Query *query, deparse_context *context,
+ bool colNamesVisible);
static void get_utility_query_def(Query *query, deparse_context *context);
static void get_basic_select_query(Query *query, deparse_context *context,
TupleDesc resultDesc, bool colNamesVisible);
@@ -5448,6 +5450,10 @@ get_query_def(Query *query, StringInfo buf, List *parentnamespace,
get_delete_query_def(query, &context, colNamesVisible);
break;
+ case CMD_MERGE:
+ get_merge_query_def(query, &context, colNamesVisible);
+ break;
+
case CMD_NOTHING:
appendStringInfoString(buf, "NOTHING");
break;
@@ -7043,6 +7049,119 @@ get_delete_query_def(Query *query, deparse_context *context,
}
}
+/* ----------
+ * get_merge_query_def - Parse back a MERGE parsetree
+ * ----------
+ */
+static void
+get_merge_query_def(Query *query, deparse_context *context,
+ bool colNamesVisible)
+{
+
+ StringInfo buf = context->buf;
+ RangeTblEntry *rte;
+ ListCell *lc;
+
+ /* Insert the WITH clause if given */
+ get_with_clause(query, context);
+
+ /*
+ * Start the query with MERGE INTO relname
+ */
+ rte = rt_fetch(query->resultRelation, query->rtable);
+ Assert(rte->rtekind == RTE_RELATION);
+ if (PRETTY_INDENT(context))
+ {
+ appendStringInfoChar(buf, ' ');
+ context->indentLevel += PRETTYINDENT_STD;
+ }
+ appendStringInfo(buf, "MERGE INTO %s%s",
+ only_marker(rte),
+ generate_relation_name(rte->relid, NIL));
+
+ /* Print the relation alias, if needed */
+ get_rte_alias(rte, query->resultRelation, false, context);
+
+ get_from_clause(query, " USING ", context);
+ appendContextKeyword(context, " ON ",
+ -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+ get_rule_expr(query->jointree->quals, context, false);
+
+ appendStringInfoChar(buf, '\n'); /* FIXME proper way to do this? */
+
+ /* FIXME missing: OVERRIDING clause */
+ foreach(lc, query->mergeActionList)
+ {
+ MergeAction *action = lfirst_node(MergeAction, lc);
+
+ appendStringInfo(buf, "WHEN %sMATCHED", action->matched ? "" : "NOT ");
+
+ if (action->qual)
+ {
+ appendContextKeyword(context, " AND ",
+ -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+ get_rule_expr(action->qual, context, false);
+ }
+ appendContextKeyword(context, " THEN ", -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+
+ if (action->commandType == CMD_INSERT)
+ {
+ ListCell *lc2;
+ char *sep = "";
+ List *strippedexprs = NIL;
+
+ appendStringInfoString(buf, "INSERT ");
+
+ /*
+ * This matches what get_insert_query_def does for the VALUES
+ * clause
+ */
+ if (action->targetList)
+ appendStringInfoChar(buf, '(');
+ foreach(lc2, action->targetList)
+ {
+ TargetEntry *tle = (TargetEntry *) lfirst(lc2);
+
+ Assert(!tle->resjunk);
+
+ appendStringInfoString(buf, sep);
+ sep = ", ";
+
+ appendStringInfoString(buf, quote_identifier(get_attname(rte->relid,
+ tle->resno,
+ false)));
+ strippedexprs = lappend(strippedexprs,
+ processIndirection((Node *) tle->expr,
+ context));
+ }
+ if (action->targetList)
+ appendStringInfoString(buf, ") ");
+ if (strippedexprs)
+ {
+ appendContextKeyword(context, "VALUES (",
+ -PRETTYINDENT_STD, PRETTYINDENT_STD, 2);
+ get_rule_list_toplevel(strippedexprs, context, false);
+ appendStringInfoChar(buf, ')');
+ }
+ else
+ appendStringInfoString(buf, "DEFAULT VALUES");
+ }
+ else if (action->commandType == CMD_UPDATE)
+ {
+ appendStringInfoString(buf, "UPDATE SET ");
+ get_update_query_targetlist_def(query, action->targetList, context, rte);
+ }
+ else if (action->commandType == CMD_DELETE)
+ appendStringInfoString(buf, "DELETE ");
+ else if (action->commandType == CMD_NOTHING)
+ appendStringInfoString(buf, "DO NOTHING ");
+
+ appendStringInfoChar(buf, '\n');
+ }
+
+ /* No RETURNING support yet */
+ Assert(query->returningList == NIL);
+}
/* ----------
* get_utility_query_def - Parse back a UTILITY parsetree
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
index 133d42117c0..2e71dda4228 100644
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -1319,6 +1319,74 @@ ERROR: syntax error at or near "RETURNING"
LINE 10: RETURNING *;
^
ROLLBACK;
+-- Verify that ruleutils can properly deparse
+CREATE TABLE sf_target(id int, data text, filling int[]);
+CREATE TABLE sf_source(id int, data text);
+CREATE OR REPLACE FUNCTION public.merge_sf_test()
+ RETURNS void
+ LANGUAGE sql
+BEGIN ATOMIC
+ MERGE INTO sf_target
+ USING sf_source s
+ ON (s.id = sf_target.id)
+WHEN MATCHED
+ AND ((s.id + sf_target.id) = 42)
+ THEN UPDATE SET data = (repeat(sf_target.data, s.id) || s.data), id = length(s.data)
+WHEN NOT MATCHED
+ AND (s.data IS NOT NULL)
+ THEN INSERT (data, id)
+ VALUES (s.data, s.id)
+WHEN MATCHED
+ AND (length((s.data || sf_target.data)) > 10)
+ THEN UPDATE SET data = s.data
+WHEN MATCHED
+ THEN UPDATE SET filling[s.id] = sf_target.id
+WHEN NOT MATCHED
+ AND (s.id > 200)
+ THEN INSERT DEFAULT VALUES
+WHEN NOT MATCHED
+ AND (s.id > 100)
+ THEN INSERT (id, data)
+ VALUES (s.id, DEFAULT)
+WHEN NOT MATCHED
+ THEN INSERT (filling[1], id)
+ VALUES (s.id, s.id)
+;
+END;
+\sf+ merge_sf_test
+ CREATE OR REPLACE FUNCTION public.merge_sf_test()
+ RETURNS void
+ LANGUAGE sql
+1 BEGIN ATOMIC
+2 MERGE INTO sf_target
+3 USING sf_source s
+4 ON (s.id = sf_target.id)
+5 WHEN MATCHED
+6 AND ((s.id + sf_target.id) = 42)
+7 THEN UPDATE SET data = (repeat(sf_target.data, s.id) || s.data), id = length(s.data)
+8 WHEN NOT MATCHED
+9 AND (s.data IS NOT NULL)
+10 THEN INSERT (data, id)
+11 VALUES (s.data, s.id)
+12 WHEN MATCHED
+13 AND (length((s.data || sf_target.data)) > 10)
+14 THEN UPDATE SET data = s.data
+15 WHEN MATCHED
+16 THEN UPDATE SET filling[s.id] = sf_target.id
+17 WHEN NOT MATCHED
+18 AND (s.id > 200)
+19 THEN INSERT DEFAULT VALUES
+20 WHEN NOT MATCHED
+21 AND (s.id > 100)
+22 THEN INSERT (id, data)
+23 VALUES (s.id, DEFAULT)
+24 WHEN NOT MATCHED
+25 THEN INSERT (filling[1], id)
+26 VALUES (s.id, s.id)
+27 ;
+28 END
+DROP FUNCTION merge_sf_test;
+DROP TABLE sf_target, sf_source;
-- EXPLAIN
CREATE TABLE ex_mtarget (a int, b int)
WITH (autovacuum_enabled=off);
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
index 4cf6db908b5..41662089905 100644
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -870,6 +870,44 @@ WHEN MATCHED AND tid < 2 THEN
RETURNING *;
ROLLBACK;
+-- Verify that ruleutils can properly deparse
+CREATE TABLE sf_target(id int, data text, filling int[]);
+CREATE TABLE sf_source(id int, data text);
+CREATE OR REPLACE FUNCTION public.merge_sf_test()
+ RETURNS void
+ LANGUAGE sql
+BEGIN ATOMIC
+ MERGE INTO sf_target
+ USING sf_source s
+ ON (s.id = sf_target.id)
+WHEN MATCHED
+ AND ((s.id + sf_target.id) = 42)
+ THEN UPDATE SET data = (repeat(sf_target.data, s.id) || s.data), id = length(s.data)
+WHEN NOT MATCHED
+ AND (s.data IS NOT NULL)
+ THEN INSERT (data, id)
+ VALUES (s.data, s.id)
+WHEN MATCHED
+ AND (length((s.data || sf_target.data)) > 10)
+ THEN UPDATE SET data = s.data
+WHEN MATCHED
+ THEN UPDATE SET filling[s.id] = sf_target.id
+WHEN NOT MATCHED
+ AND (s.id > 200)
+ THEN INSERT DEFAULT VALUES
+WHEN NOT MATCHED
+ AND (s.id > 100)
+ THEN INSERT (id, data)
+ VALUES (s.id, DEFAULT)
+WHEN NOT MATCHED
+ THEN INSERT (filling[1], id)
+ VALUES (s.id, s.id)
+;
+END;
+\sf+ merge_sf_test
+DROP FUNCTION merge_sf_test;
+DROP TABLE sf_target, sf_source;
+
-- EXPLAIN
CREATE TABLE ex_mtarget (a int, b int)
WITH (autovacuum_enabled=off);
--
2.39.2
--3dv3bnnc7qew3llg--
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: [CAUTION!! freemail] Re: Partial aggregates pushdown
@ 2024-01-27 01:57 vignesh C <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: vignesh C @ 2024-01-27 01:57 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Robert Haas <[email protected]>; pgsql-hackers; Tom Lane <[email protected]>; Alexander Pyhalov <[email protected]>; Bruce Momjian <[email protected]>; Stephen Frost <[email protected]>; Finnerty, Jim <[email protected]>; Andres Freund <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>
On Thu, 7 Dec 2023 at 05:41, [email protected]
<[email protected]> wrote:
>
> Hi Mr.Haas.
>
> > -----Original Message-----
> > From: Robert Haas <[email protected]>
> > Sent: Wednesday, December 6, 2023 10:25 PM
> > On Wed, Dec 6, 2023 at 3:41 AM [email protected]
> > <[email protected]> wrote:
> > > Are you concerned about the hassle and potential human errors of
> > > manually adding new partial aggregation functions, rather than the catalog becoming bloated?
> >
> > I'm concerned about both.
> Understood. Thank you for your response.
>
> > > The process of creating partial aggregation functions from aggregation
> > > functions can be automated, so I believe this issue can be resolved.
> > > However, automating it may increase the size of the patch even more, so overall, approach#2 might be better.
> > > To implement approach #2, it would be necessary to investigate how much additional code is required.
> >
> > Yes. Unfortunately I fear that there is quite a lot of work left to do here in order to produce a committable feature. To me it
> > seems necessary to conduct an investigation of approach #2. If the result of that investigation is that nothing major
> > stands in the way of approach #2, then I think we should adopt it, which is more work. In addition, the problems around
> > transmitting serialized bytea blobs between machines that can't be assumed to fully trust each other will need to be
> > addressed in some way, which seems like it will require a good deal of design work, forming some kind of consensus, and
> > then implementation work to follow. In addition to that there may be some small problems that need to be solved at a
> > detail level, such as the HAVING issue. I think the last category won't be too hard to sort out, but that still leaves two really
> > major areas to address.
> Yes, I agree with you. It is clear that further investigation and discussion are still needed.
> I would be grateful if we can resolve this issue gradually. I would also like to continue the discussion if possible in the future.
Thanks for all the efforts on this patch. I have changed the status of
the commitfest entry to "Returned with Feedback" as there is still
some work to get this patch out. Feel free to continue the discussion
and add a new entry when the patch is in a reviewable shape.
Regards,
Vignesh
^ permalink raw reply [nested|flat] 42+ messages in thread
* RE: Partial aggregates pushdown
@ 2025-03-28 02:00 [email protected] <[email protected]>
2025-03-29 22:22 ` Re: Partial aggregates pushdown Bruce Momjian <[email protected]>
0 siblings, 1 reply; 42+ messages in thread
From: [email protected] @ 2025-03-28 02:00 UTC (permalink / raw)
To: Bruce Momjian <[email protected]>; Tomas Vondra <[email protected]>; +Cc: Jelte Fennema-Nio <[email protected]>; pgsql-hackers; Robert Haas <[email protected]>; Tom Lane <[email protected]>; Stephen Frost <[email protected]>; Andres Freund <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; vignesh C <[email protected]>; Alexander Pyhalov <[email protected]>
Hi Bruce, Jelte, hackers.
I apologize for my late response.
> From: Jelte Fennema-Nio <[email protected]>
> Sent: Thursday, August 8, 2024 8:49 PM
> SUMMARY OF THREAD
>
> The design of patch 0001 is agreed upon by everyone on the thread (so far). This adds the PARTIAL_AGGREGATE label for
> aggregates, which will cause the finalfunc not to run. It also starts using PARTIAL_AGGREGATE for pushdown of
> aggregates in postgres_fdw. In 0001 PARTIAL_AGGREGATE is only supported for aggregates with a non-internal/pseudo
> type as the stype.
>
> The design for patch 0002 is still under debate. This would expand on the functionality added by adding support for
> PARTIAL_AGGREGATE for aggregates with an internal stype. This is done by returning a byte array containing the bytes
> that the serialfunc of the aggregate returns.
>
> A competing proposal for 0002 is to instead change aggregates to not use an internal stype anymore, and create dedicated
> types. The main downside here is that infunc and outfunc would need to be added for text serialization, in addition to the
> binary serialization. An open question is: Can we change the requirements for CREATE TYPE, so that types can be created
> without infunc and outfunc.
I rebased the patch 0001 and add the documentation to it.
Best regards, Yuki Fujii
--
Yuki Fujii
Information Technology R&D Center, Mitsubishi Electric Corporation
> -----Original Message-----
> From: Bruce Momjian <[email protected]>
> Sent: Friday, August 23, 2024 4:31 AM
> To: Tomas Vondra <[email protected]>
> Cc: Jelte Fennema-Nio <[email protected]>; Fujii Yuki/藤井 雄規(MELCO/情報総研 ＤＭ最適Ｇ)
> <[email protected]>; PostgreSQL-development <[email protected]>; Robert Haas
> <[email protected]>; Tom Lane <[email protected]>; Stephen Frost <[email protected]>; Andres Freund
> <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>;
> Daniel Gustafsson <[email protected]>; vignesh C <[email protected]>; Alexander Pyhalov
> <[email protected]>
> Subject: Re: Partial aggregates pushdown
>
> On Thu, Aug 22, 2024 at 08:31:11PM +0200, Tomas Vondra wrote:
> > > My question is related to #3 and #4. For #3, if we are going to be
> > > building infrastructure to handle passing int128 for AVG, wouldn't
> > > it be wiser to create an int128 type and an int128 array type, and
> > > then use method #2 to handle those, rather than creating custom code
> > > just to read/write int128 values for FDWs aggregate passing alone.
> > >
> >
> > Yep, adding int128 as a data type would extend this to aggregates that
> > store state as int128 (or array of int128).
>
> Great, I am not too far off then.
>
> > > For #4, can we use or improve the RECORD data type to handle #4 ---
> > > that seems preferable to creating custom FDWs aggregate passing code.
> > >
> > > I know the open question was whether we should create custom FDWs
> > > aggregate passing functions or custom data types for FDWs aggregate
> > > passing, but I am asking if we can improve existing facilities, like
> > > int128 or record passing, to reduce the need for some of these.
> > >
> >
> > But which code would produce the record? AFAIK it can't happen in some
> > generic executor code, because that only sees "internal" for each
> > aggregate. The exact structure of the aggstate is private within the
> > code of each aggregate - the record would have to be built there, no?
> >
> > I imagine we'd add this for each aggregate as a new pair of functions
> > to build/parse the record, but that's kinda the serial/deserial way we
> > discussed earlier.
> >
> > Or are you suggesting we'd actually say:
> >
> > CREATE AGGREGATE xyz(...) (
> > STYPE = record,
> > ...
> > )
>
> So my idea from the email I just sent is to create a pg_proc.proargtypes-like column (data type oidvector) for pg_aggregate
> which stores the oids of the values we want to return, so AVG(interval) would have an array of the oids for interval and int8,
> e.g.:
>
> SELECT oid FROM pg_type WHERE typname = 'interval';
> oid
> ------
> 1186
>
> SELECT oid FROM pg_type WHERE typname = 'int8';
> oid
> -----
> 20
>
> SELECT '1186 20'::oidvector;
> oidvector
> -----------
> 1186 20
>
> It seems all four methods could use this, again assuming we create
> int128/int16 and whatever other types we need.
>
> --
> Bruce Momjian <[email protected]> https://momjian.us
> EDB https://enterprisedb.com
>
> Only you can decide what is important to you.
Attachments:
[application/octet-stream] 0001-Partial-Aggregate-Pushdown-Add-PARTIAL_AGGREGATE-key.patch (113.8K, ../../TYRPR01MB13941CEA16574771B1BFD130595A02@TYRPR01MB13941.jpnprd01.prod.outlook.com/2-0001-Partial-Aggregate-Pushdown-Add-PARTIAL_AGGREGATE-key.patch)
download | inline diff:
From ac871aa602a8330af87c7c9cef3e918b65e60030 Mon Sep 17 00:00:00 2001
From: Yuki Fujii <[email protected]>
Date: Mon, 17 Mar 2025 09:20:58 -0400
Subject: [PATCH] Partial Aggregate Pushdown; Add PARTIAL_AGGREGATE keyword
---
contrib/postgres_fdw/deparse.c | 75 +-
.../postgres_fdw/expected/postgres_fdw.out | 638 ++++++++++++++++--
contrib/postgres_fdw/option.c | 4 +-
contrib/postgres_fdw/postgres_fdw.c | 43 +-
contrib/postgres_fdw/postgres_fdw.h | 12 +-
contrib/postgres_fdw/sql/postgres_fdw.sql | 112 ++-
doc/src/sgml/postgres-fdw.sgml | 103 ++-
doc/src/sgml/syntax.sgml | 27 +-
src/backend/executor/nodeAgg.c | 6 +-
src/backend/nodes/makefuncs.c | 1 +
src/backend/optimizer/plan/planner.c | 133 +++-
src/backend/optimizer/plan/setrefs.c | 1 +
src/backend/optimizer/prep/prepagg.c | 1 +
src/backend/parser/gram.y | 18 +-
src/backend/parser/parse_agg.c | 21 +-
src/backend/parser/parse_clause.c | 1 +
src/backend/parser/parse_expr.c | 3 +-
src/backend/parser/parse_func.c | 24 +-
src/backend/utils/adt/ruleutils.c | 3 +-
src/include/nodes/parsenodes.h | 1 +
src/include/nodes/pathnodes.h | 5 +
src/include/nodes/primnodes.h | 3 +
src/include/parser/kwlist.h | 1 +
src/include/parser/parse_agg.h | 2 +-
src/test/regress/expected/aggregates.out | 8 +
.../regress/expected/partition_aggregate.out | 29 +
src/test/regress/sql/aggregates.sql | 4 +
src/test/regress/sql/partition_aggregate.sql | 5 +
28 files changed, 1149 insertions(+), 135 deletions(-)
diff --git a/contrib/postgres_fdw/deparse.c b/contrib/postgres_fdw/deparse.c
index 6e7dc3d2df9..e30bf8c4329 100644
--- a/contrib/postgres_fdw/deparse.c
+++ b/contrib/postgres_fdw/deparse.c
@@ -204,7 +204,7 @@ static bool is_subquery_var(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
static void get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
int *relno, int *colno);
-
+static bool partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo);
/*
* Examine each qual clause in input_conds, and classify them into two groups,
@@ -909,8 +909,9 @@ foreign_expr_walker(Node *node,
if (!IS_UPPER_REL(glob_cxt->foreignrel))
return false;
- /* Only non-split aggregates are pushable. */
- if (agg->aggsplit != AGGSPLIT_SIMPLE)
+ if (agg->aggsplit != AGGSPLIT_SIMPLE && agg->aggsplit != AGGSPLIT_INITIAL_SERIAL)
+ return false;
+ if (agg->aggsplit == AGGSPLIT_INITIAL_SERIAL && !partial_agg_ok(agg, fpinfo))
return false;
/* As usual, it must be shippable. */
@@ -3657,18 +3658,21 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
StringInfo buf = context->buf;
bool use_variadic;
- /* Only basic, non-split aggregation accepted. */
- Assert(node->aggsplit == AGGSPLIT_SIMPLE);
+
+ Assert((node->aggsplit == AGGSPLIT_SIMPLE) ||
+ (node->aggsplit == AGGSPLIT_INITIAL_SERIAL));
/* Check if need to print VARIADIC (cf. ruleutils.c) */
use_variadic = node->aggvariadic;
/* Find aggregate name from aggfnoid which is a pg_proc entry */
appendFunctionName(node->aggfnoid, context);
+
appendStringInfoChar(buf, '(');
- /* Add DISTINCT */
- appendStringInfoString(buf, (node->aggdistinct != NIL) ? "DISTINCT " : "");
+ /* Add DISTINCT or PARTIAL_AGGREGATE */
+ if (node->aggdistinct != NIL)
+ appendStringInfoString(buf, "DISTINCT ");
if (AGGKIND_IS_ORDERED_SET(node->aggkind))
{
@@ -3738,6 +3742,9 @@ deparseAggref(Aggref *node, deparse_expr_cxt *context)
}
appendStringInfoChar(buf, ')');
+
+ if (node->aggsplit == AGGSPLIT_INITIAL_SERIAL)
+ appendStringInfoString(buf, " PARTIAL_AGGREGATE ");
}
/*
@@ -3863,9 +3870,10 @@ appendGroupByClause(List *tlist, deparse_expr_cxt *context)
Query *query = context->root->parse;
ListCell *lc;
bool first = true;
+ PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) context->foreignrel->fdw_private;
/* Nothing to be done, if there's no GROUP BY clause in the query. */
- if (!query->groupClause)
+ if (!fpinfo->group_clause)
return;
appendStringInfoString(buf, " GROUP BY ");
@@ -3883,7 +3891,7 @@ appendGroupByClause(List *tlist, deparse_expr_cxt *context)
* to empty, and in any case the redundancy situation on the remote might
* be different than what we think here.
*/
- foreach(lc, query->groupClause)
+ foreach(lc, fpinfo->group_clause)
{
SortGroupClause *grp = (SortGroupClause *) lfirst(lc);
@@ -4204,3 +4212,52 @@ get_relation_column_alias_ids(Var *node, RelOptInfo *foreignrel,
/* Shouldn't get here */
elog(ERROR, "unexpected expression in subquery output");
}
+
+/*
+ * ----------
+ * Check that partial aggregate "agg" is safe to push down.
+ *
+ * It is pushdown-safe when all of the following conditions are true:
+ *
+ * * agg is an AGGKIND_NORMAL aggregate which contains no DISTINCT or
+ * ORDER BY clauses
+ * * remote server can return partial aggregate results
+ * ----------
+ */
+static bool
+partial_agg_ok(Aggref *agg, PgFdwRelationInfo *fpinfo)
+{
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+ bool partial_agg_ok = true;
+
+ Assert(agg->aggsplit == AGGSPLIT_INITIAL_SERIAL);
+
+ /* We don't support complex partial aggregates */
+ if (agg->aggdistinct || agg->aggkind != AGGKIND_NORMAL || agg->aggorder != NIL)
+ return false;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+
+ partial_agg_ok = fpinfo->partial_aggregate_support;
+ if(aggform->aggtranstype == INTERNALOID ||
+ get_typtype(aggform->aggtranstype) == TYPTYPE_PSEUDO)
+ partial_agg_ok = false;
+ if(partial_agg_ok){
+ if (fpinfo->remoteversion == 0)
+ {
+ PGconn *conn = GetConnection(fpinfo->user, false, NULL);
+
+ fpinfo->remoteversion = PQserverVersion(conn);
+ }
+
+ if (fpinfo->remoteversion != PG_VERSION_NUM)
+ partial_agg_ok = false;
+ }
+
+ ReleaseSysCache(aggtup);
+ return partial_agg_ok;
+}
diff --git a/contrib/postgres_fdw/expected/postgres_fdw.out b/contrib/postgres_fdw/expected/postgres_fdw.out
index bb4ed3059c4..f99d93c1805 100644
--- a/contrib/postgres_fdw/expected/postgres_fdw.out
+++ b/contrib/postgres_fdw/expected/postgres_fdw.out
@@ -10260,36 +10260,57 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.1');
+CREATE TYPE mood AS enum ('sad', 'ok', 'happy');
+ALTER EXTENSION postgres_fdw ADD TYPE mood;
+CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4,
+ c_int4array _int4, c_interval interval,
+ c_money money, c_1c text, c_1b bytea,
+ c_bit bit(2), c_1or3int2 int2,
+ c_1or3int4 int4, c_1or3int8 int8,
+ c_bool bool, c_enum mood, c_pg_lsn pg_lsn,
+ c_tid tid, c_int4range int4range,
+ c_int4multirange int4multirange,
+ c_time time, c_timetz timetz,
+ c_timestamp timestamp, c_timestamptz timestamptz,
+ c_xid8 xid8)
+ PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+SET TimeZone = 'UTC';
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO (20) SERVER loopback OPTIONS (table_name 'pagg_tab_p2');
CREATE FOREIGN TABLE fpagg_tab_p3 PARTITION OF pagg_tab FOR VALUES FROM (20) TO (30) SERVER loopback OPTIONS (table_name 'pagg_tab_p3');
ANALYZE pagg_tab;
+ANALYZE pagg_tab_p1;
+ANALYZE pagg_tab_p2;
+ANALYZE pagg_tab_p3;
ANALYZE fpagg_tab_p1;
ANALYZE fpagg_tab_p2;
ANALYZE fpagg_tab_p3;
+SET extra_float_digits = 0;
-- When GROUP BY clause matches with PARTITION KEY.
-- Plan with partitionwise aggregates is disabled
SET enable_partitionwise_aggregate TO false;
EXPLAIN (COSTS OFF)
SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
------------------------------------------------------
- GroupAggregate
- Group Key: pagg_tab.a
- Filter: (avg(pagg_tab.b) < '22'::numeric)
- -> Append
- -> Foreign Scan on fpagg_tab_p1 pagg_tab_1
- -> Foreign Scan on fpagg_tab_p2 pagg_tab_2
- -> Foreign Scan on fpagg_tab_p3 pagg_tab_3
-(7 rows)
+ QUERY PLAN
+-----------------------------------------------------------
+ Sort
+ Sort Key: pagg_tab.a
+ -> HashAggregate
+ Group Key: pagg_tab.a
+ Filter: (avg(pagg_tab.b) < '22'::numeric)
+ -> Append
+ -> Foreign Scan on fpagg_tab_p1 pagg_tab_1
+ -> Foreign Scan on fpagg_tab_p2 pagg_tab_2
+ -> Foreign Scan on fpagg_tab_p3 pagg_tab_3
+(9 rows)
-- Plan with partitionwise aggregates is enabled
SET enable_partitionwise_aggregate TO true;
@@ -10323,32 +10344,34 @@ SELECT a, sum(b), min(b), count(*) FROM pagg_tab GROUP BY a HAVING avg(b) < 22 O
-- Should have all the columns in the target list for the given relation
EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
- QUERY PLAN
---------------------------------------------------------------------------------------------
- Merge Append
+ QUERY PLAN
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: t1.a, (count(((t1.*)::pagg_tab)))
Sort Key: t1.a
- -> GroupAggregate
- Output: t1.a, count(((t1.*)::pagg_tab))
- Group Key: t1.a
- Filter: (avg(t1.b) < '22'::numeric)
- -> Foreign Scan on public.fpagg_tab_p1 t1
- Output: t1.a, t1.*, t1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p1 ORDER BY a ASC NULLS LAST
- -> GroupAggregate
- Output: t1_1.a, count(((t1_1.*)::pagg_tab))
- Group Key: t1_1.a
- Filter: (avg(t1_1.b) < '22'::numeric)
- -> Foreign Scan on public.fpagg_tab_p2 t1_1
- Output: t1_1.a, t1_1.*, t1_1.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p2 ORDER BY a ASC NULLS LAST
- -> GroupAggregate
- Output: t1_2.a, count(((t1_2.*)::pagg_tab))
- Group Key: t1_2.a
- Filter: (avg(t1_2.b) < '22'::numeric)
- -> Foreign Scan on public.fpagg_tab_p3 t1_2
- Output: t1_2.a, t1_2.*, t1_2.b
- Remote SQL: SELECT a, b, c FROM public.pagg_tab_p3 ORDER BY a ASC NULLS LAST
-(23 rows)
+ -> Append
+ -> HashAggregate
+ Output: t1.a, count(((t1.*)::pagg_tab))
+ Group Key: t1.a
+ Filter: (avg(t1.b) < '22'::numeric)
+ -> Foreign Scan on public.fpagg_tab_p1 t1
+ Output: t1.a, t1.*, t1.b
+ Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p1
+ -> HashAggregate
+ Output: t1_1.a, count(((t1_1.*)::pagg_tab))
+ Group Key: t1_1.a
+ Filter: (avg(t1_1.b) < '22'::numeric)
+ -> Foreign Scan on public.fpagg_tab_p2 t1_1
+ Output: t1_1.a, t1_1.*, t1_1.b
+ Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p2
+ -> HashAggregate
+ Output: t1_2.a, count(((t1_2.*)::pagg_tab))
+ Group Key: t1_2.a
+ Filter: (avg(t1_2.b) < '22'::numeric)
+ -> Foreign Scan on public.fpagg_tab_p3 t1_2
+ Output: t1_2.a, t1_2.*, t1_2.b
+ Remote SQL: SELECT a, b, c, c_serial, c_int4array, c_interval, c_money, c_1c, c_1b, c_bit, c_1or3int2, c_1or3int4, c_1or3int8, c_bool, c_enum, c_pg_lsn, c_tid, c_int4range, c_int4multirange, c_time, c_timetz, c_timestamp, c_timestamptz, c_xid8 FROM public.pagg_tab_p3
+(25 rows)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
a | count
@@ -10361,27 +10384,534 @@ SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
21 | 100
(6 rows)
--- When GROUP BY clause does not match with PARTITION KEY.
-EXPLAIN (COSTS OFF)
+-- Partial aggregates are safe to push down when there is a HAVING clause
+EXPLAIN (VERBOSE, COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
- QUERY PLAN
------------------------------------------------------------
+ QUERY PLAN
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ Filter: (sum(pagg_tab.a) < 700)
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab.a))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE , sum(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_1.a))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE , sum(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), (PARTIAL sum(pagg_tab_2.a))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE , sum(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 GROUP BY 1
+(20 rows)
+
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+ b | avg | max | count
+----+---------------------+-----+-------
+ 0 | 10.0000000000000000 | 20 | 60
+ 1 | 11.0000000000000000 | 21 | 60
+ 10 | 10.0000000000000000 | 20 | 60
+ 11 | 11.0000000000000000 | 21 | 60
+ 20 | 10.0000000000000000 | 20 | 60
+ 21 | 11.0000000000000000 | 21 | 60
+ 30 | 10.0000000000000000 | 20 | 60
+ 31 | 11.0000000000000000 | 21 | 60
+ 40 | 10.0000000000000000 | 20 | 60
+ 41 | 11.0000000000000000 | 21 | 60
+(10 rows)
+
+-- Partial aggregates are safe to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, avg(pagg_tab.a), max(pagg_tab.a), count(*)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | avg | max | count
+----+---------------------+-----+-------
+ 0 | 10.0000000000000000 | 20 | 60
+ 1 | 11.0000000000000000 | 21 | 60
+ 2 | 12.0000000000000000 | 22 | 60
+ 3 | 13.0000000000000000 | 23 | 60
+ 4 | 14.0000000000000000 | 24 | 60
+ 5 | 15.0000000000000000 | 25 | 60
+ 6 | 16.0000000000000000 | 26 | 60
+ 7 | 17.0000000000000000 | 27 | 60
+ 8 | 18.0000000000000000 | 28 | 60
+ 9 | 19.0000000000000000 | 29 | 60
+ 10 | 10.0000000000000000 | 20 | 60
+ 11 | 11.0000000000000000 | 21 | 60
+ 12 | 12.0000000000000000 | 22 | 60
+ 13 | 13.0000000000000000 | 23 | 60
+ 14 | 14.0000000000000000 | 24 | 60
+ 15 | 15.0000000000000000 | 25 | 60
+ 16 | 16.0000000000000000 | 26 | 60
+ 17 | 17.0000000000000000 | 27 | 60
+ 18 | 18.0000000000000000 | 28 | 60
+ 19 | 19.0000000000000000 | 29 | 60
+ 20 | 10.0000000000000000 | 20 | 60
+ 21 | 11.0000000000000000 | 21 | 60
+ 22 | 12.0000000000000000 | 22 | 60
+ 23 | 13.0000000000000000 | 23 | 60
+ 24 | 14.0000000000000000 | 24 | 60
+ 25 | 15.0000000000000000 | 25 | 60
+ 26 | 16.0000000000000000 | 26 | 60
+ 27 | 17.0000000000000000 | 27 | 60
+ 28 | 18.0000000000000000 | 28 | 60
+ 29 | 19.0000000000000000 | 29 | 60
+ 30 | 10.0000000000000000 | 20 | 60
+ 31 | 11.0000000000000000 | 21 | 60
+ 32 | 12.0000000000000000 | 22 | 60
+ 33 | 13.0000000000000000 | 23 | 60
+ 34 | 14.0000000000000000 | 24 | 60
+ 35 | 15.0000000000000000 | 25 | 60
+ 36 | 16.0000000000000000 | 26 | 60
+ 37 | 17.0000000000000000 | 27 | 60
+ 38 | 18.0000000000000000 | 28 | 60
+ 39 | 19.0000000000000000 | 29 | 60
+ 40 | 10.0000000000000000 | 20 | 60
+ 41 | 11.0000000000000000 | 21 | 60
+ 42 | 12.0000000000000000 | 22 | 60
+ 43 | 13.0000000000000000 | 23 | 60
+ 44 | 14.0000000000000000 | 24 | 60
+ 45 | 15.0000000000000000 | 25 | 60
+ 46 | 16.0000000000000000 | 26 | 60
+ 47 | 17.0000000000000000 | 27 | 60
+ 48 | 18.0000000000000000 | 28 | 60
+ 49 | 19.0000000000000000 | 29 | 60
+(50 rows)
+
+-- Partial aggregates are safe to push down even if we need both variable and variable-based expression
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(a), max(a), count(*), (b/2)::numeric FROM pagg_tab GROUP BY b/2 ORDER BY 4;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Sort
+ Output: (avg(pagg_tab.a)), (max(pagg_tab.a)), (count(*)), ((((pagg_tab.b / 2)))::numeric), ((pagg_tab.b / 2))
+ Sort Key: ((((pagg_tab.b / 2)))::numeric)
+ -> Finalize HashAggregate
+ Output: avg(pagg_tab.a), max(pagg_tab.a), count(*), (((pagg_tab.b / 2)))::numeric, ((pagg_tab.b / 2))
+ Group Key: ((pagg_tab.b / 2))
+ -> Append
+ -> Foreign Scan
+ Output: ((pagg_tab.b / 2)), (PARTIAL avg(pagg_tab.a)), (PARTIAL max(pagg_tab.a)), (PARTIAL count(*)), pagg_tab.b
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 GROUP BY 1, 2
+ -> Foreign Scan
+ Output: ((pagg_tab_1.b / 2)), (PARTIAL avg(pagg_tab_1.a)), (PARTIAL max(pagg_tab_1.a)), (PARTIAL count(*)), pagg_tab_1.b
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 GROUP BY 1, 2
+ -> Foreign Scan
+ Output: ((pagg_tab_2.b / 2)), (PARTIAL avg(pagg_tab_2.a)), (PARTIAL max(pagg_tab_2.a)), (PARTIAL count(*)), pagg_tab_2.b
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE , max(a) PARTIAL_AGGREGATE , count(*) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 GROUP BY 1, 2
+(19 rows)
+
+SELECT avg(a), max(a), count(*), (b/2)::numeric FROM pagg_tab GROUP BY b/2 ORDER BY 4;
+ avg | max | count | numeric
+---------------------+-----+-------+---------
+ 10.5000000000000000 | 21 | 120 | 0
+ 12.5000000000000000 | 23 | 120 | 1
+ 14.5000000000000000 | 25 | 120 | 2
+ 16.5000000000000000 | 27 | 120 | 3
+ 18.5000000000000000 | 29 | 120 | 4
+ 10.5000000000000000 | 21 | 120 | 5
+ 12.5000000000000000 | 23 | 120 | 6
+ 14.5000000000000000 | 25 | 120 | 7
+ 16.5000000000000000 | 27 | 120 | 8
+ 18.5000000000000000 | 29 | 120 | 9
+ 10.5000000000000000 | 21 | 120 | 10
+ 12.5000000000000000 | 23 | 120 | 11
+ 14.5000000000000000 | 25 | 120 | 12
+ 16.5000000000000000 | 27 | 120 | 13
+ 18.5000000000000000 | 29 | 120 | 14
+ 10.5000000000000000 | 21 | 120 | 15
+ 12.5000000000000000 | 23 | 120 | 16
+ 14.5000000000000000 | 25 | 120 | 17
+ 16.5000000000000000 | 27 | 120 | 18
+ 18.5000000000000000 | 29 | 120 | 19
+ 10.5000000000000000 | 21 | 120 | 20
+ 12.5000000000000000 | 23 | 120 | 21
+ 14.5000000000000000 | 25 | 120 | 22
+ 16.5000000000000000 | 27 | 120 | 23
+ 18.5000000000000000 | 29 | 120 | 24
+(25 rows)
+
+-- Partial aggregates pushdown behaves sane with non-shipped grouping elements
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(a), (b/2)::numeric, 'test' test FROM pagg_tab GROUP BY 3, b/2 ORDER BY 1, 2 LIMIT 10;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------
+ Limit
+ Output: (avg(pagg_tab.a)), ((((pagg_tab.b / 2)))::numeric), 'test'::text, ((pagg_tab.b / 2))
+ -> Sort
+ Output: (avg(pagg_tab.a)), ((((pagg_tab.b / 2)))::numeric), 'test'::text, ((pagg_tab.b / 2))
+ Sort Key: (avg(pagg_tab.a)), ((((pagg_tab.b / 2)))::numeric)
+ -> Finalize HashAggregate
+ Output: avg(pagg_tab.a), (((pagg_tab.b / 2)))::numeric, 'test'::text, ((pagg_tab.b / 2))
+ Group Key: ((pagg_tab.b / 2))
+ -> Append
+ -> Foreign Scan
+ Output: ((pagg_tab.b / 2)), (PARTIAL avg(pagg_tab.a)), pagg_tab.b
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 GROUP BY 1, 2
+ -> Foreign Scan
+ Output: ((pagg_tab_1.b / 2)), (PARTIAL avg(pagg_tab_1.a)), pagg_tab_1.b
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 GROUP BY 1, 2
+ -> Foreign Scan
+ Output: ((pagg_tab_2.b / 2)), (PARTIAL avg(pagg_tab_2.a)), pagg_tab_2.b
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT (b / 2), b, avg(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 GROUP BY 1, 2
+(21 rows)
+
+SELECT avg(a), (b/2)::numeric, 'test' test FROM pagg_tab GROUP BY 3, b/2 ORDER BY 1, 2 LIMIT 10;
+ avg | numeric | test
+---------------------+---------+------
+ 10.5000000000000000 | 0 | test
+ 10.5000000000000000 | 5 | test
+ 10.5000000000000000 | 10 | test
+ 10.5000000000000000 | 15 | test
+ 10.5000000000000000 | 20 | test
+ 12.5000000000000000 | 1 | test
+ 12.5000000000000000 | 6 | test
+ 12.5000000000000000 | 11 | test
+ 12.5000000000000000 | 16 | test
+ 12.5000000000000000 | 21 | test
+(10 rows)
+
+-- Partial aggregates are safe to push down for all built-in aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT
+ /* The cases that don't require import or export functions */
+ bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8),
+ bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8),
+ bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8),
+ bool_and(c_bool),
+ bool_or(c_bool),
+ every(c_bool),
+ max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_1b), max(c_xid8),
+ min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), max(c_1b), min(c_xid8),
+ avg(b::int4)
+ FROM pagg_tab WHERE c_serial between 1 and 30;
+ QUERY PLAN
+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: bit_and(pagg_tab.c_bit), bit_and(pagg_tab.c_1or3int2), bit_and(pagg_tab.c_1or3int4), bit_and(pagg_tab.c_1or3int8), bit_or(pagg_tab.c_bit), bit_or(pagg_tab.c_1or3int2), bit_or(pagg_tab.c_1or3int4), bit_or(pagg_tab.c_1or3int8), bit_xor(pagg_tab.c_bit), bit_xor(pagg_tab.c_1or3int2), bit_xor(pagg_tab.c_1or3int4), bit_xor(pagg_tab.c_1or3int8), bool_and(pagg_tab.c_bool), bool_or(pagg_tab.c_bool), every(pagg_tab.c_bool), max((pagg_tab.c_1c)::character(1)), max(('01-01-2000'::date + pagg_tab.b)), max(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), max((pagg_tab.b)::real), max((pagg_tab.b)::double precision), max((pagg_tab.b)::smallint), max(pagg_tab.b), max((pagg_tab.b)::bigint), max(pagg_tab.c_interval), max(pagg_tab.c_money), max((pagg_tab.b)::numeric), max((pagg_tab.b)::oid), max(pagg_tab.c_pg_lsn), max(pagg_tab.c_tid), max(pagg_tab.c_1c), max(pagg_tab.c_time), max(pagg_tab.c_timetz), max(pagg_tab.c_timestamp), max(pagg_tab.c_timestamptz), max(pagg_tab.c_1b), max(pagg_tab.c_xid8), min((pagg_tab.c_1c)::character(1)), min(('01-01-2000'::date + pagg_tab.b)), min(('0.0.0.0'::inet + (pagg_tab.b)::bigint)), min((pagg_tab.b)::real), min((pagg_tab.b)::double precision), min((pagg_tab.b)::smallint), min(pagg_tab.b), min((pagg_tab.b)::bigint), min(pagg_tab.c_interval), min(pagg_tab.c_money), min((pagg_tab.b)::numeric), min((pagg_tab.b)::oid), min(pagg_tab.c_pg_lsn), min(pagg_tab.c_tid), min(pagg_tab.c_1c), min(pagg_tab.c_time), min(pagg_tab.c_timetz), min(pagg_tab.c_timestamp), min(pagg_tab.c_timestamptz), max(pagg_tab.c_1b), min(pagg_tab.c_xid8), avg(pagg_tab.b)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL bit_and(pagg_tab.c_bit)), (PARTIAL bit_and(pagg_tab.c_1or3int2)), (PARTIAL bit_and(pagg_tab.c_1or3int4)), (PARTIAL bit_and(pagg_tab.c_1or3int8)), (PARTIAL bit_or(pagg_tab.c_bit)), (PARTIAL bit_or(pagg_tab.c_1or3int2)), (PARTIAL bit_or(pagg_tab.c_1or3int4)), (PARTIAL bit_or(pagg_tab.c_1or3int8)), (PARTIAL bit_xor(pagg_tab.c_bit)), (PARTIAL bit_xor(pagg_tab.c_1or3int2)), (PARTIAL bit_xor(pagg_tab.c_1or3int4)), (PARTIAL bit_xor(pagg_tab.c_1or3int8)), (PARTIAL bool_and(pagg_tab.c_bool)), (PARTIAL bool_or(pagg_tab.c_bool)), (PARTIAL every(pagg_tab.c_bool)), (PARTIAL max((pagg_tab.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL max((pagg_tab.b)::real)), (PARTIAL max((pagg_tab.b)::double precision)), (PARTIAL max((pagg_tab.b)::smallint)), (PARTIAL max(pagg_tab.b)), (PARTIAL max((pagg_tab.b)::bigint)), (PARTIAL max(pagg_tab.c_interval)), (PARTIAL max(pagg_tab.c_money)), (PARTIAL max((pagg_tab.b)::numeric)), (PARTIAL max((pagg_tab.b)::oid)), (PARTIAL max(pagg_tab.c_pg_lsn)), (PARTIAL max(pagg_tab.c_tid)), (PARTIAL max(pagg_tab.c_1c)), (PARTIAL max(pagg_tab.c_time)), (PARTIAL max(pagg_tab.c_timetz)), (PARTIAL max(pagg_tab.c_timestamp)), (PARTIAL max(pagg_tab.c_timestamptz)), (PARTIAL max(pagg_tab.c_1b)), (PARTIAL max(pagg_tab.c_xid8)), (PARTIAL min((pagg_tab.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab.b)::bigint))), (PARTIAL min((pagg_tab.b)::real)), (PARTIAL min((pagg_tab.b)::double precision)), (PARTIAL min((pagg_tab.b)::smallint)), (PARTIAL min(pagg_tab.b)), (PARTIAL min((pagg_tab.b)::bigint)), (PARTIAL min(pagg_tab.c_interval)), (PARTIAL min(pagg_tab.c_money)), (PARTIAL min((pagg_tab.b)::numeric)), (PARTIAL min((pagg_tab.b)::oid)), (PARTIAL min(pagg_tab.c_pg_lsn)), (PARTIAL min(pagg_tab.c_tid)), (PARTIAL min(pagg_tab.c_1c)), (PARTIAL min(pagg_tab.c_time)), (PARTIAL min(pagg_tab.c_timetz)), (PARTIAL min(pagg_tab.c_timestamp)), (PARTIAL min(pagg_tab.c_timestamptz)), (PARTIAL min(pagg_tab.c_xid8)), (PARTIAL avg(pagg_tab.b))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT bit_and(c_bit) PARTIAL_AGGREGATE , bit_and(c_1or3int2) PARTIAL_AGGREGATE , bit_and(c_1or3int4) PARTIAL_AGGREGATE , bit_and(c_1or3int8) PARTIAL_AGGREGATE , bit_or(c_bit) PARTIAL_AGGREGATE , bit_or(c_1or3int2) PARTIAL_AGGREGATE , bit_or(c_1or3int4) PARTIAL_AGGREGATE , bit_or(c_1or3int8) PARTIAL_AGGREGATE , bit_xor(c_bit) PARTIAL_AGGREGATE , bit_xor(c_1or3int2) PARTIAL_AGGREGATE , bit_xor(c_1or3int4) PARTIAL_AGGREGATE , bit_xor(c_1or3int8) PARTIAL_AGGREGATE , bool_and(c_bool) PARTIAL_AGGREGATE , bool_or(c_bool) PARTIAL_AGGREGATE , every(c_bool) PARTIAL_AGGREGATE , max(c_1c::character(1)) PARTIAL_AGGREGATE , max(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , max(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , max(b::real) PARTIAL_AGGREGATE , max(b::double precision) PARTIAL_AGGREGATE , max(b::smallint) PARTIAL_AGGREGATE , max(b) PARTIAL_AGGREGATE , max(b::bigint) PARTIAL_AGGREGATE , max(c_interval) PARTIAL_AGGREGATE , max(c_money) PARTIAL_AGGREGATE , max(b::numeric) PARTIAL_AGGREGATE , max(b::oid) PARTIAL_AGGREGATE , max(c_pg_lsn) PARTIAL_AGGREGATE , max(c_tid) PARTIAL_AGGREGATE , max(c_1c) PARTIAL_AGGREGATE , max(c_time) PARTIAL_AGGREGATE , max(c_timetz) PARTIAL_AGGREGATE , max(c_timestamp) PARTIAL_AGGREGATE , max(c_timestamptz) PARTIAL_AGGREGATE , max(c_1b) PARTIAL_AGGREGATE , max(c_xid8) PARTIAL_AGGREGATE , min(c_1c::character(1)) PARTIAL_AGGREGATE , min(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , min(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , min(b::real) PARTIAL_AGGREGATE , min(b::double precision) PARTIAL_AGGREGATE , min(b::smallint) PARTIAL_AGGREGATE , min(b) PARTIAL_AGGREGATE , min(b::bigint) PARTIAL_AGGREGATE , min(c_interval) PARTIAL_AGGREGATE , min(c_money) PARTIAL_AGGREGATE , min(b::numeric) PARTIAL_AGGREGATE , min(b::oid) PARTIAL_AGGREGATE , min(c_pg_lsn) PARTIAL_AGGREGATE , min(c_tid) PARTIAL_AGGREGATE , min(c_1c) PARTIAL_AGGREGATE , min(c_time) PARTIAL_AGGREGATE , min(c_timetz) PARTIAL_AGGREGATE , min(c_timestamp) PARTIAL_AGGREGATE , min(c_timestamptz) PARTIAL_AGGREGATE , min(c_xid8) PARTIAL_AGGREGATE , avg(b) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+ -> Foreign Scan
+ Output: (PARTIAL bit_and(pagg_tab_1.c_bit)), (PARTIAL bit_and(pagg_tab_1.c_1or3int2)), (PARTIAL bit_and(pagg_tab_1.c_1or3int4)), (PARTIAL bit_and(pagg_tab_1.c_1or3int8)), (PARTIAL bit_or(pagg_tab_1.c_bit)), (PARTIAL bit_or(pagg_tab_1.c_1or3int2)), (PARTIAL bit_or(pagg_tab_1.c_1or3int4)), (PARTIAL bit_or(pagg_tab_1.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_1.c_bit)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_1.c_1or3int8)), (PARTIAL bool_and(pagg_tab_1.c_bool)), (PARTIAL bool_or(pagg_tab_1.c_bool)), (PARTIAL every(pagg_tab_1.c_bool)), (PARTIAL max((pagg_tab_1.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL max((pagg_tab_1.b)::real)), (PARTIAL max((pagg_tab_1.b)::double precision)), (PARTIAL max((pagg_tab_1.b)::smallint)), (PARTIAL max(pagg_tab_1.b)), (PARTIAL max((pagg_tab_1.b)::bigint)), (PARTIAL max(pagg_tab_1.c_interval)), (PARTIAL max(pagg_tab_1.c_money)), (PARTIAL max((pagg_tab_1.b)::numeric)), (PARTIAL max((pagg_tab_1.b)::oid)), (PARTIAL max(pagg_tab_1.c_pg_lsn)), (PARTIAL max(pagg_tab_1.c_tid)), (PARTIAL max(pagg_tab_1.c_1c)), (PARTIAL max(pagg_tab_1.c_time)), (PARTIAL max(pagg_tab_1.c_timetz)), (PARTIAL max(pagg_tab_1.c_timestamp)), (PARTIAL max(pagg_tab_1.c_timestamptz)), (PARTIAL max(pagg_tab_1.c_1b)), (PARTIAL max(pagg_tab_1.c_xid8)), (PARTIAL min((pagg_tab_1.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_1.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_1.b)::bigint))), (PARTIAL min((pagg_tab_1.b)::real)), (PARTIAL min((pagg_tab_1.b)::double precision)), (PARTIAL min((pagg_tab_1.b)::smallint)), (PARTIAL min(pagg_tab_1.b)), (PARTIAL min((pagg_tab_1.b)::bigint)), (PARTIAL min(pagg_tab_1.c_interval)), (PARTIAL min(pagg_tab_1.c_money)), (PARTIAL min((pagg_tab_1.b)::numeric)), (PARTIAL min((pagg_tab_1.b)::oid)), (PARTIAL min(pagg_tab_1.c_pg_lsn)), (PARTIAL min(pagg_tab_1.c_tid)), (PARTIAL min(pagg_tab_1.c_1c)), (PARTIAL min(pagg_tab_1.c_time)), (PARTIAL min(pagg_tab_1.c_timetz)), (PARTIAL min(pagg_tab_1.c_timestamp)), (PARTIAL min(pagg_tab_1.c_timestamptz)), (PARTIAL min(pagg_tab_1.c_xid8)), (PARTIAL avg(pagg_tab_1.b))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT bit_and(c_bit) PARTIAL_AGGREGATE , bit_and(c_1or3int2) PARTIAL_AGGREGATE , bit_and(c_1or3int4) PARTIAL_AGGREGATE , bit_and(c_1or3int8) PARTIAL_AGGREGATE , bit_or(c_bit) PARTIAL_AGGREGATE , bit_or(c_1or3int2) PARTIAL_AGGREGATE , bit_or(c_1or3int4) PARTIAL_AGGREGATE , bit_or(c_1or3int8) PARTIAL_AGGREGATE , bit_xor(c_bit) PARTIAL_AGGREGATE , bit_xor(c_1or3int2) PARTIAL_AGGREGATE , bit_xor(c_1or3int4) PARTIAL_AGGREGATE , bit_xor(c_1or3int8) PARTIAL_AGGREGATE , bool_and(c_bool) PARTIAL_AGGREGATE , bool_or(c_bool) PARTIAL_AGGREGATE , every(c_bool) PARTIAL_AGGREGATE , max(c_1c::character(1)) PARTIAL_AGGREGATE , max(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , max(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , max(b::real) PARTIAL_AGGREGATE , max(b::double precision) PARTIAL_AGGREGATE , max(b::smallint) PARTIAL_AGGREGATE , max(b) PARTIAL_AGGREGATE , max(b::bigint) PARTIAL_AGGREGATE , max(c_interval) PARTIAL_AGGREGATE , max(c_money) PARTIAL_AGGREGATE , max(b::numeric) PARTIAL_AGGREGATE , max(b::oid) PARTIAL_AGGREGATE , max(c_pg_lsn) PARTIAL_AGGREGATE , max(c_tid) PARTIAL_AGGREGATE , max(c_1c) PARTIAL_AGGREGATE , max(c_time) PARTIAL_AGGREGATE , max(c_timetz) PARTIAL_AGGREGATE , max(c_timestamp) PARTIAL_AGGREGATE , max(c_timestamptz) PARTIAL_AGGREGATE , max(c_1b) PARTIAL_AGGREGATE , max(c_xid8) PARTIAL_AGGREGATE , min(c_1c::character(1)) PARTIAL_AGGREGATE , min(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , min(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , min(b::real) PARTIAL_AGGREGATE , min(b::double precision) PARTIAL_AGGREGATE , min(b::smallint) PARTIAL_AGGREGATE , min(b) PARTIAL_AGGREGATE , min(b::bigint) PARTIAL_AGGREGATE , min(c_interval) PARTIAL_AGGREGATE , min(c_money) PARTIAL_AGGREGATE , min(b::numeric) PARTIAL_AGGREGATE , min(b::oid) PARTIAL_AGGREGATE , min(c_pg_lsn) PARTIAL_AGGREGATE , min(c_tid) PARTIAL_AGGREGATE , min(c_1c) PARTIAL_AGGREGATE , min(c_time) PARTIAL_AGGREGATE , min(c_timetz) PARTIAL_AGGREGATE , min(c_timestamp) PARTIAL_AGGREGATE , min(c_timestamptz) PARTIAL_AGGREGATE , min(c_xid8) PARTIAL_AGGREGATE , avg(b) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+ -> Foreign Scan
+ Output: (PARTIAL bit_and(pagg_tab_2.c_bit)), (PARTIAL bit_and(pagg_tab_2.c_1or3int2)), (PARTIAL bit_and(pagg_tab_2.c_1or3int4)), (PARTIAL bit_and(pagg_tab_2.c_1or3int8)), (PARTIAL bit_or(pagg_tab_2.c_bit)), (PARTIAL bit_or(pagg_tab_2.c_1or3int2)), (PARTIAL bit_or(pagg_tab_2.c_1or3int4)), (PARTIAL bit_or(pagg_tab_2.c_1or3int8)), (PARTIAL bit_xor(pagg_tab_2.c_bit)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int2)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int4)), (PARTIAL bit_xor(pagg_tab_2.c_1or3int8)), (PARTIAL bool_and(pagg_tab_2.c_bool)), (PARTIAL bool_or(pagg_tab_2.c_bool)), (PARTIAL every(pagg_tab_2.c_bool)), (PARTIAL max((pagg_tab_2.c_1c)::character(1))), (PARTIAL max(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL max(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL max((pagg_tab_2.b)::real)), (PARTIAL max((pagg_tab_2.b)::double precision)), (PARTIAL max((pagg_tab_2.b)::smallint)), (PARTIAL max(pagg_tab_2.b)), (PARTIAL max((pagg_tab_2.b)::bigint)), (PARTIAL max(pagg_tab_2.c_interval)), (PARTIAL max(pagg_tab_2.c_money)), (PARTIAL max((pagg_tab_2.b)::numeric)), (PARTIAL max((pagg_tab_2.b)::oid)), (PARTIAL max(pagg_tab_2.c_pg_lsn)), (PARTIAL max(pagg_tab_2.c_tid)), (PARTIAL max(pagg_tab_2.c_1c)), (PARTIAL max(pagg_tab_2.c_time)), (PARTIAL max(pagg_tab_2.c_timetz)), (PARTIAL max(pagg_tab_2.c_timestamp)), (PARTIAL max(pagg_tab_2.c_timestamptz)), (PARTIAL max(pagg_tab_2.c_1b)), (PARTIAL max(pagg_tab_2.c_xid8)), (PARTIAL min((pagg_tab_2.c_1c)::character(1))), (PARTIAL min(('01-01-2000'::date + pagg_tab_2.b))), (PARTIAL min(('0.0.0.0'::inet + (pagg_tab_2.b)::bigint))), (PARTIAL min((pagg_tab_2.b)::real)), (PARTIAL min((pagg_tab_2.b)::double precision)), (PARTIAL min((pagg_tab_2.b)::smallint)), (PARTIAL min(pagg_tab_2.b)), (PARTIAL min((pagg_tab_2.b)::bigint)), (PARTIAL min(pagg_tab_2.c_interval)), (PARTIAL min(pagg_tab_2.c_money)), (PARTIAL min((pagg_tab_2.b)::numeric)), (PARTIAL min((pagg_tab_2.b)::oid)), (PARTIAL min(pagg_tab_2.c_pg_lsn)), (PARTIAL min(pagg_tab_2.c_tid)), (PARTIAL min(pagg_tab_2.c_1c)), (PARTIAL min(pagg_tab_2.c_time)), (PARTIAL min(pagg_tab_2.c_timetz)), (PARTIAL min(pagg_tab_2.c_timestamp)), (PARTIAL min(pagg_tab_2.c_timestamptz)), (PARTIAL min(pagg_tab_2.c_xid8)), (PARTIAL avg(pagg_tab_2.b))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT bit_and(c_bit) PARTIAL_AGGREGATE , bit_and(c_1or3int2) PARTIAL_AGGREGATE , bit_and(c_1or3int4) PARTIAL_AGGREGATE , bit_and(c_1or3int8) PARTIAL_AGGREGATE , bit_or(c_bit) PARTIAL_AGGREGATE , bit_or(c_1or3int2) PARTIAL_AGGREGATE , bit_or(c_1or3int4) PARTIAL_AGGREGATE , bit_or(c_1or3int8) PARTIAL_AGGREGATE , bit_xor(c_bit) PARTIAL_AGGREGATE , bit_xor(c_1or3int2) PARTIAL_AGGREGATE , bit_xor(c_1or3int4) PARTIAL_AGGREGATE , bit_xor(c_1or3int8) PARTIAL_AGGREGATE , bool_and(c_bool) PARTIAL_AGGREGATE , bool_or(c_bool) PARTIAL_AGGREGATE , every(c_bool) PARTIAL_AGGREGATE , max(c_1c::character(1)) PARTIAL_AGGREGATE , max(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , max(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , max(b::real) PARTIAL_AGGREGATE , max(b::double precision) PARTIAL_AGGREGATE , max(b::smallint) PARTIAL_AGGREGATE , max(b) PARTIAL_AGGREGATE , max(b::bigint) PARTIAL_AGGREGATE , max(c_interval) PARTIAL_AGGREGATE , max(c_money) PARTIAL_AGGREGATE , max(b::numeric) PARTIAL_AGGREGATE , max(b::oid) PARTIAL_AGGREGATE , max(c_pg_lsn) PARTIAL_AGGREGATE , max(c_tid) PARTIAL_AGGREGATE , max(c_1c) PARTIAL_AGGREGATE , max(c_time) PARTIAL_AGGREGATE , max(c_timetz) PARTIAL_AGGREGATE , max(c_timestamp) PARTIAL_AGGREGATE , max(c_timestamptz) PARTIAL_AGGREGATE , max(c_1b) PARTIAL_AGGREGATE , max(c_xid8) PARTIAL_AGGREGATE , min(c_1c::character(1)) PARTIAL_AGGREGATE , min(('01-01-2000'::date + b)) PARTIAL_AGGREGATE , min(('0.0.0.0'::inet + b)) PARTIAL_AGGREGATE , min(b::real) PARTIAL_AGGREGATE , min(b::double precision) PARTIAL_AGGREGATE , min(b::smallint) PARTIAL_AGGREGATE , min(b) PARTIAL_AGGREGATE , min(b::bigint) PARTIAL_AGGREGATE , min(c_interval) PARTIAL_AGGREGATE , min(c_money) PARTIAL_AGGREGATE , min(b::numeric) PARTIAL_AGGREGATE , min(b::oid) PARTIAL_AGGREGATE , min(c_pg_lsn) PARTIAL_AGGREGATE , min(c_tid) PARTIAL_AGGREGATE , min(c_1c) PARTIAL_AGGREGATE , min(c_time) PARTIAL_AGGREGATE , min(c_timetz) PARTIAL_AGGREGATE , min(c_timestamp) PARTIAL_AGGREGATE , min(c_timestamptz) PARTIAL_AGGREGATE , min(c_xid8) PARTIAL_AGGREGATE , avg(b) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+(15 rows)
+
+SELECT
+ /* The cases that don't require import or export functions */
+ bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8),
+ bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8),
+ bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8),
+ bool_and(c_bool),
+ bool_or(c_bool),
+ every(c_bool),
+ max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_1b), max(c_xid8),
+ min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), max(c_1b), min(c_xid8),
+ avg(b::int4)
+ FROM pagg_tab WHERE c_serial between 1 and 30;
+ bit_and | bit_and | bit_and | bit_and | bit_or | bit_or | bit_or | bit_or | bit_xor | bit_xor | bit_xor | bit_xor | bool_and | bool_or | every | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | max | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | min | max | min | avg
+---------+---------+---------+---------+--------+--------+--------+--------+---------+---------+---------+---------+----------+---------+-------+-----+------------+----------+-----+-----+-----+-----+-----+---------+-------+-----+-----+------+--------+-----+----------+-------------+--------------------------+------------------------------+------+-----+-----+------------+---------+-----+-----+-----+-----+-----+-----+-------+-----+-----+-----+-------+-----+----------+-------------+--------------------------+------------------------------+------+-----+---------------------
+ 01 | 1 | 1 | 1 | 11 | 3 | 3 | 3 | 10 | 2 | 2 | 2 | f | t | f | 9 | 01-31-2000 | 0.0.0.30 | 30 | 30 | 30 | 30 | 30 | @ 1 sec | $1.00 | 30 | 30 | 0/30 | (0,30) | 9 | 00:00:30 | 00:00:30+00 | Sat Jan 01 00:00:30 2000 | Sat Jan 01 00:00:30 2000 UTC | \x39 | 9 | 0 | 01-02-2000 | 0.0.0.1 | 1 | 1 | 1 | 1 | 1 | @ 0 | $0.00 | 1 | 1 | 0/1 | (0,1) | 0 | 00:00:01 | 00:00:01+00 | Sat Jan 01 00:00:01 2000 | Sat Jan 01 00:00:01 2000 UTC | \x39 | 0 | 15.5000000000000000
+(1 row)
+
+-- Tests for partial_aggregate_support
+ALTER SERVER loopback OPTIONS (ADD partial_aggregate_support 'false');
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+---------------------------------------------------------------------------
Finalize GroupAggregate
+ Output: pagg_tab.b, max(pagg_tab.a)
Group Key: pagg_tab.b
- Filter: (sum(pagg_tab.a) < 700)
+ -> Sort
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a))
+ Sort Key: pagg_tab.b
+ -> Append
+ -> Partial HashAggregate
+ Output: pagg_tab.b, PARTIAL max(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab
+ Output: pagg_tab.b, pagg_tab.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1
+ -> Partial HashAggregate
+ Output: pagg_tab_1.b, PARTIAL max(pagg_tab_1.a)
+ Group Key: pagg_tab_1.b
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.b, pagg_tab_1.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2
+ -> Partial HashAggregate
+ Output: pagg_tab_2.b, PARTIAL max(pagg_tab_2.a)
+ Group Key: pagg_tab_2.b
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.b, pagg_tab_2.a
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3
+(25 rows)
+
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max
+----+-----
+ 0 | 20
+ 1 | 21
+ 2 | 22
+ 3 | 23
+ 4 | 24
+ 5 | 25
+ 6 | 26
+ 7 | 27
+ 8 | 28
+ 9 | 29
+ 10 | 20
+ 11 | 21
+ 12 | 22
+ 13 | 23
+ 14 | 24
+ 15 | 25
+ 16 | 26
+ 17 | 27
+ 18 | 28
+ 19 | 29
+ 20 | 20
+ 21 | 21
+ 22 | 22
+ 23 | 23
+ 24 | 24
+ 25 | 25
+ 26 | 26
+ 27 | 27
+ 28 | 28
+ 29 | 29
+ 30 | 20
+ 31 | 21
+ 32 | 22
+ 33 | 23
+ 34 | 24
+ 35 | 25
+ 36 | 26
+ 37 | 27
+ 38 | 28
+ 39 | 29
+ 40 | 20
+ 41 | 21
+ 42 | 22
+ 43 | 23
+ 44 | 24
+ 45 | 25
+ 46 | 26
+ 47 | 27
+ 48 | 28
+ 49 | 29
+(50 rows)
+
+ALTER SERVER loopback OPTIONS (SET partial_aggregate_support 'true');
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------
+ Sort
+ Output: pagg_tab.b, (max(pagg_tab.a))
+ Sort Key: pagg_tab.b
+ -> Finalize HashAggregate
+ Output: pagg_tab.b, max(pagg_tab.a)
+ Group Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan
+ Output: pagg_tab.b, (PARTIAL max(pagg_tab.a))
+ Relations: Aggregate on (public.fpagg_tab_p1 pagg_tab)
+ Remote SQL: SELECT b, max(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p1 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_1.b, (PARTIAL max(pagg_tab_1.a))
+ Relations: Aggregate on (public.fpagg_tab_p2 pagg_tab_1)
+ Remote SQL: SELECT b, max(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p2 GROUP BY 1
+ -> Foreign Scan
+ Output: pagg_tab_2.b, (PARTIAL max(pagg_tab_2.a))
+ Relations: Aggregate on (public.fpagg_tab_p3 pagg_tab_2)
+ Remote SQL: SELECT b, max(a) PARTIAL_AGGREGATE FROM public.pagg_tab_p3 GROUP BY 1
+(19 rows)
+
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+ b | max
+----+-----
+ 0 | 20
+ 1 | 21
+ 2 | 22
+ 3 | 23
+ 4 | 24
+ 5 | 25
+ 6 | 26
+ 7 | 27
+ 8 | 28
+ 9 | 29
+ 10 | 20
+ 11 | 21
+ 12 | 22
+ 13 | 23
+ 14 | 24
+ 15 | 25
+ 16 | 26
+ 17 | 27
+ 18 | 28
+ 19 | 29
+ 20 | 20
+ 21 | 21
+ 22 | 22
+ 23 | 23
+ 24 | 24
+ 25 | 25
+ 26 | 26
+ 27 | 27
+ 28 | 28
+ 29 | 29
+ 30 | 20
+ 31 | 21
+ 32 | 22
+ 33 | 23
+ 34 | 24
+ 35 | 25
+ 36 | 26
+ 37 | 27
+ 38 | 28
+ 39 | 29
+ 40 | 20
+ 41 | 21
+ 42 | 22
+ 43 | 23
+ 44 | 24
+ 45 | 25
+ 46 | 26
+ 47 | 27
+ 48 | 28
+ 49 | 29
+(50 rows)
+
+ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '1.0');
+SET enable_partitionwise_join=on;
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(t1.b) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a);
+ QUERY PLAN
+--------------------------------------------------------------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: max(t1.b)
+ -> Append
+ -> Foreign Scan
+ Output: (PARTIAL max(t1.b))
+ Relations: Aggregate on ((public.fpagg_tab_p1 t1) INNER JOIN (public.fpagg_tab_p1 t2))
+ Remote SQL: SELECT max(r4.b) PARTIAL_AGGREGATE FROM (public.pagg_tab_p1 r4 INNER JOIN public.pagg_tab_p1 r7 ON (((r4.a = r7.a))))
+ -> Foreign Scan
+ Output: (PARTIAL max(t1_1.b))
+ Relations: Aggregate on ((public.fpagg_tab_p2 t1_1) INNER JOIN (public.fpagg_tab_p2 t2_1))
+ Remote SQL: SELECT max(r5.b) PARTIAL_AGGREGATE FROM (public.pagg_tab_p2 r5 INNER JOIN public.pagg_tab_p2 r8 ON (((r5.a = r8.a))))
+ -> Foreign Scan
+ Output: (PARTIAL max(t1_2.b))
+ Relations: Aggregate on ((public.fpagg_tab_p3 t1_2) INNER JOIN (public.fpagg_tab_p3 t2_2))
+ Remote SQL: SELECT max(r6.b) PARTIAL_AGGREGATE FROM (public.pagg_tab_p3 r6 INNER JOIN public.pagg_tab_p3 r9 ON (((r6.a = r9.a))))
+(15 rows)
+
+SELECT max(t1.b) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a);
+ max
+-----
+ 49
+(1 row)
+
+RESET enable_partitionwise_join;
+ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '0.1');
+ALTER SERVER loopback OPTIONS (DROP partial_aggregate_support);
+-- It is unsafe to push down partial aggregates which contain DISTINCT clauses
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------
+ Aggregate
+ Output: max(pagg_tab.a), count(DISTINCT pagg_tab.b)
-> Merge Append
Sort Key: pagg_tab.b
- -> Partial GroupAggregate
- Group Key: pagg_tab.b
- -> Foreign Scan on fpagg_tab_p1 pagg_tab
- -> Partial GroupAggregate
- Group Key: pagg_tab_1.b
- -> Foreign Scan on fpagg_tab_p2 pagg_tab_1
- -> Partial GroupAggregate
- Group Key: pagg_tab_2.b
- -> Foreign Scan on fpagg_tab_p3 pagg_tab_2
-(14 rows)
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.a, pagg_tab_1.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p1 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.a, pagg_tab_2.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p2 ORDER BY b ASC NULLS LAST
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.a, pagg_tab_3.b
+ Remote SQL: SELECT a, b FROM public.pagg_tab_p3 ORDER BY b ASC NULLS LAST
+(13 rows)
+
+SELECT max(a), count(distinct b) FROM pagg_tab;
+ max | count
+-----+-------
+ 29 | 50
+(1 row)
+
+-- It is unsafe to push down partial aggregates which contain ORDER BY clauses
+EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30;
+ QUERY PLAN
+-----------------------------------------------------------------------------------------------------------------
+ Aggregate
+ Output: array_agg(pagg_tab.b ORDER BY pagg_tab.b)
+ -> Sort
+ Output: pagg_tab.b
+ Sort Key: pagg_tab.b
+ -> Append
+ -> Foreign Scan on public.fpagg_tab_p1 pagg_tab_1
+ Output: pagg_tab_1.b
+ Remote SQL: SELECT b FROM public.pagg_tab_p1 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+ -> Foreign Scan on public.fpagg_tab_p2 pagg_tab_2
+ Output: pagg_tab_2.b
+ Remote SQL: SELECT b FROM public.pagg_tab_p2 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+ -> Foreign Scan on public.fpagg_tab_p3 pagg_tab_3
+ Output: pagg_tab_3.b
+ Remote SQL: SELECT b FROM public.pagg_tab_p3 WHERE ((c_serial >= 1)) AND ((c_serial <= 30))
+(15 rows)
+
+SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30;
+ array_agg
+------------------------------------------------------------------------------------
+ {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30}
+(1 row)
+ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost);
+RESET TimeZone;
-- ===================================================================
-- access rights and superuser
-- ===================================================================
diff --git a/contrib/postgres_fdw/option.c b/contrib/postgres_fdw/option.c
index d0766f007d2..d3aad176942 100644
--- a/contrib/postgres_fdw/option.c
+++ b/contrib/postgres_fdw/option.c
@@ -125,7 +125,8 @@ postgres_fdw_validator(PG_FUNCTION_ARGS)
strcmp(def->defname, "async_capable") == 0 ||
strcmp(def->defname, "parallel_commit") == 0 ||
strcmp(def->defname, "parallel_abort") == 0 ||
- strcmp(def->defname, "keep_connections") == 0)
+ strcmp(def->defname, "keep_connections") == 0 ||
+ strcmp(def->defname, "partial_aggregate_support") == 0)
{
/* these accept only boolean values */
(void) defGetBoolean(def);
@@ -267,6 +268,7 @@ InitPgFdwOptions(void)
/* batch_size is available on both server and table */
{"batch_size", ForeignServerRelationId, false},
{"batch_size", ForeignTableRelationId, false},
+ {"partial_aggregate_support", ForeignServerRelationId, false},
/* async_capable is available on both server and table */
{"async_capable", ForeignServerRelationId, false},
{"async_capable", ForeignTableRelationId, false},
diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 1131a8bf77e..bf75f347fc0 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -18,6 +18,8 @@
#include "access/sysattr.h"
#include "access/table.h"
#include "catalog/pg_opfamily.h"
+#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "commands/defrem.h"
#include "commands/explain.h"
#include "commands/explain_format.h"
@@ -34,6 +36,7 @@
#include "optimizer/pathnode.h"
#include "optimizer/paths.h"
#include "optimizer/planmain.h"
+#include "optimizer/planner.h"
#include "optimizer/prep.h"
#include "optimizer/restrictinfo.h"
#include "optimizer/tlist.h"
@@ -48,6 +51,7 @@
#include "utils/rel.h"
#include "utils/sampling.h"
#include "utils/selfuncs.h"
+#include "utils/syscache.h"
PG_MODULE_MAGIC;
@@ -519,7 +523,7 @@ static bool foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel,
JoinType jointype, RelOptInfo *outerrel, RelOptInfo *innerrel,
JoinPathExtraData *extra);
static bool foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual);
+ GroupPathExtraData *extra);
static List *get_useful_pathkeys_for_relation(PlannerInfo *root,
RelOptInfo *rel);
static List *get_useful_ecs_for_relation(PlannerInfo *root, RelOptInfo *rel);
@@ -650,6 +654,8 @@ postgresGetForeignRelSize(PlannerInfo *root,
fpinfo->fdw_tuple_cost = DEFAULT_FDW_TUPLE_COST;
fpinfo->shippable_extensions = NIL;
fpinfo->fetch_size = 100;
+ fpinfo->remoteversion = 0;
+ fpinfo->partial_aggregate_support = true;
fpinfo->async_capable = false;
apply_server_options(fpinfo);
@@ -661,7 +667,7 @@ postgresGetForeignRelSize(PlannerInfo *root,
* should match what ExecCheckPermissions() does. If we fail due to lack
* of permissions, the query would have failed at runtime anyway.
*/
- if (fpinfo->use_remote_estimate)
+ if (fpinfo->use_remote_estimate || fpinfo->partial_aggregate_support)
{
Oid userid;
@@ -6052,9 +6058,9 @@ foreign_join_ok(PlannerInfo *root, RelOptInfo *joinrel, JoinType jointype,
fpinfo->pushdown_safe = true;
/* Get user mapping */
- if (fpinfo->use_remote_estimate)
+ if (fpinfo->use_remote_estimate || fpinfo->partial_aggregate_support)
{
- if (fpinfo_o->use_remote_estimate)
+ if (fpinfo_o->use_remote_estimate || fpinfo_o->partial_aggregate_support)
fpinfo->user = fpinfo_o->user;
else
fpinfo->user = fpinfo_i->user;
@@ -6229,6 +6235,8 @@ apply_server_options(PgFdwRelationInfo *fpinfo)
if (strcmp(def->defname, "use_remote_estimate") == 0)
fpinfo->use_remote_estimate = defGetBoolean(def);
+ else if (strcmp(def->defname, "partial_aggregate_support") == 0)
+ fpinfo->partial_aggregate_support = defGetBoolean(def);
else if (strcmp(def->defname, "fdw_startup_cost") == 0)
(void) parse_real(defGetString(def), &fpinfo->fdw_startup_cost, 0,
NULL);
@@ -6299,6 +6307,8 @@ merge_fdw_options(PgFdwRelationInfo *fpinfo,
fpinfo->shippable_extensions = fpinfo_o->shippable_extensions;
fpinfo->use_remote_estimate = fpinfo_o->use_remote_estimate;
fpinfo->fetch_size = fpinfo_o->fetch_size;
+ fpinfo->remoteversion = fpinfo_o->remoteversion;
+ fpinfo->partial_aggregate_support = fpinfo_o->partial_aggregate_support;
fpinfo->async_capable = fpinfo_o->async_capable;
/* Merge the table level options from either side of the join. */
@@ -6485,7 +6495,7 @@ postgresGetForeignJoinPaths(PlannerInfo *root,
*/
static bool
foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
- Node *havingQual)
+ GroupPathExtraData *extra)
{
Query *query = root->parse;
PgFdwRelationInfo *fpinfo = (PgFdwRelationInfo *) grouped_rel->fdw_private;
@@ -6494,6 +6504,7 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
ListCell *lc;
int i;
List *tlist = NIL;
+ bool partial = extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL;
/* We currently don't support pushing Grouping Sets. */
if (query->groupingSets)
@@ -6527,6 +6538,12 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
* a node, as long as it's not at top level; then no match is possible.
*/
i = 0;
+ fpinfo->group_clause = query->groupClause;
+ if (partial)
+ {
+ fpinfo->group_clause = extra->groupClausePartial;
+ grouping_target = extra->partial_target;
+ }
foreach(lc, grouping_target->exprs)
{
Expr *expr = (Expr *) lfirst(lc);
@@ -6538,7 +6555,7 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
* check the whole GROUP BY clause not just processed_groupClause,
* because we will ship all of it, cf. appendGroupByClause.
*/
- if (sgref && get_sortgroupref_clause_noerr(sgref, query->groupClause))
+ if (sgref && get_sortgroupref_clause_noerr(sgref, fpinfo->group_clause))
{
TargetEntry *tle;
@@ -6624,9 +6641,9 @@ foreign_grouping_ok(PlannerInfo *root, RelOptInfo *grouped_rel,
* Classify the pushable and non-pushable HAVING clauses and save them in
* remote_conds and local_conds of the grouped rel's fpinfo.
*/
- if (havingQual)
+ if (extra->havingQual && !partial)
{
- foreach(lc, (List *) havingQual)
+ foreach(lc, (List *) extra->havingQual)
{
Expr *expr = (Expr *) lfirst(lc);
RestrictInfo *rinfo;
@@ -6740,6 +6757,7 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
/* Ignore stages we don't support; and skip any duplicate calls. */
if ((stage != UPPERREL_GROUP_AGG &&
+ stage != UPPERREL_PARTIAL_GROUP_AGG &&
stage != UPPERREL_ORDERED &&
stage != UPPERREL_FINAL) ||
output_rel->fdw_private)
@@ -6756,6 +6774,10 @@ postgresGetForeignUpperPaths(PlannerInfo *root, UpperRelationKind stage,
add_foreign_grouping_paths(root, input_rel, output_rel,
(GroupPathExtraData *) extra);
break;
+ case UPPERREL_PARTIAL_GROUP_AGG:
+ add_foreign_grouping_paths(root, input_rel, output_rel,
+ (GroupPathExtraData *) extra);
+ break;
case UPPERREL_ORDERED:
add_foreign_ordered_paths(root, input_rel, output_rel);
break;
@@ -6797,7 +6819,8 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
return;
Assert(extra->patype == PARTITIONWISE_AGGREGATE_NONE ||
- extra->patype == PARTITIONWISE_AGGREGATE_FULL);
+ extra->patype == PARTITIONWISE_AGGREGATE_FULL ||
+ extra->patype == PARTITIONWISE_AGGREGATE_PARTIAL);
/* save the input_rel as outerrel in fpinfo */
fpinfo->outerrel = input_rel;
@@ -6817,7 +6840,7 @@ add_foreign_grouping_paths(PlannerInfo *root, RelOptInfo *input_rel,
* Use HAVING qual from extra. In case of child partition, it will have
* translated Vars.
*/
- if (!foreign_grouping_ok(root, grouped_rel, extra->havingQual))
+ if (!foreign_grouping_ok(root, grouped_rel, extra))
return;
/*
diff --git a/contrib/postgres_fdw/postgres_fdw.h b/contrib/postgres_fdw/postgres_fdw.h
index 81358f3bde7..a9874b07c35 100644
--- a/contrib/postgres_fdw/postgres_fdw.h
+++ b/contrib/postgres_fdw/postgres_fdw.h
@@ -85,7 +85,16 @@ typedef struct PgFdwRelationInfo
/* Cached catalog information. */
ForeignTable *table;
ForeignServer *server;
- UserMapping *user; /* only set in use_remote_estimate mode */
+ UserMapping *user; /* only set in use_remote_estimate/partial_aggregate_support mode */
+
+ /* for partial aggregate pushdown */
+ bool partial_aggregate_support;
+
+ /*
+ * If remoteversion is zero, it means the remote server version has not
+ * been acquired.
+ */
+ int remoteversion;
int fetch_size; /* fetch size for this remote table */
@@ -111,6 +120,7 @@ typedef struct PgFdwRelationInfo
/* Grouping information */
List *grouped_tlist;
+ List *group_clause;
/* Subquery information */
bool make_outerrel_subquery; /* do we deparse outerrel as a
diff --git a/contrib/postgres_fdw/sql/postgres_fdw.sql b/contrib/postgres_fdw/sql/postgres_fdw.sql
index d45e9f8ab52..5901081c138 100644
--- a/contrib/postgres_fdw/sql/postgres_fdw.sql
+++ b/contrib/postgres_fdw/sql/postgres_fdw.sql
@@ -3201,16 +3201,33 @@ RESET enable_partitionwise_join;
-- ===================================================================
-- test partitionwise aggregates
-- ===================================================================
-
-CREATE TABLE pagg_tab (a int, b int, c text) PARTITION BY RANGE(a);
+ALTER SERVER loopback OPTIONS (ADD fdw_tuple_cost '0.1');
+
+CREATE TYPE mood AS enum ('sad', 'ok', 'happy');
+ALTER EXTENSION postgres_fdw ADD TYPE mood;
+
+CREATE TABLE pagg_tab (a int, b int, c text, c_serial int4,
+ c_int4array _int4, c_interval interval,
+ c_money money, c_1c text, c_1b bytea,
+ c_bit bit(2), c_1or3int2 int2,
+ c_1or3int4 int4, c_1or3int8 int8,
+ c_bool bool, c_enum mood, c_pg_lsn pg_lsn,
+ c_tid tid, c_int4range int4range,
+ c_int4multirange int4multirange,
+ c_time time, c_timetz timetz,
+ c_timestamp timestamp, c_timestamptz timestamptz,
+ c_xid8 xid8)
+ PARTITION BY RANGE(a);
CREATE TABLE pagg_tab_p1 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p2 (LIKE pagg_tab);
CREATE TABLE pagg_tab_p3 (LIKE pagg_tab);
-INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
-INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
-INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000') FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
+SET TimeZone = 'UTC';
+
+INSERT INTO pagg_tab_p1 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 10;
+INSERT INTO pagg_tab_p2 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 20 and (i % 30) >= 10;
+INSERT INTO pagg_tab_p3 SELECT i % 30, i % 50, to_char(i/30, 'FM0000'), i, array[(i % 2), 0], ((i % 2) || ' seconds')::interval, (i % 2)::money, ((i - 1) % 10)::text, (((i - 1) % 10)::text)::bytea, case when (i % 2) = 0 then B'01' else B'11' end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then 1 else 3 end, case when (i % 2) = 0 then true else false end, (case when (i % 2) = 0 then 'sad' else 'happy' end)::mood, ('0/' || i)::pg_lsn, ('(0,' || i || ')')::tid, int4range(0, i), int4multirange(int4range(0, i), int4range(100, 100 + i)), '00:00:00'::time + (i || ' seconds')::interval, '00:00:00'::timetz + (i || ' seconds')::interval, '2000-01-01'::timestamp + (i || ' seconds')::interval, '2000-01-01'::timestamptz + (i || ' seconds')::interval, ((i % 10)::text)::xid8 FROM generate_series(1, 3000) i WHERE (i % 30) < 30 and (i % 30) >= 20;
-- Create foreign partitions
CREATE FOREIGN TABLE fpagg_tab_p1 PARTITION OF pagg_tab FOR VALUES FROM (0) TO (10) SERVER loopback OPTIONS (table_name 'pagg_tab_p1');
@@ -3218,9 +3235,13 @@ CREATE FOREIGN TABLE fpagg_tab_p2 PARTITION OF pagg_tab FOR VALUES FROM (10) TO
CREATE FOREIGN TABLE fpagg_tab_p3 PARTITION OF pagg_tab FOR VALUES FROM (20) TO (30) SERVER loopback OPTIONS (table_name 'pagg_tab_p3');
ANALYZE pagg_tab;
+ANALYZE pagg_tab_p1;
+ANALYZE pagg_tab_p2;
+ANALYZE pagg_tab_p3;
ANALYZE fpagg_tab_p1;
ANALYZE fpagg_tab_p2;
ANALYZE fpagg_tab_p3;
+SET extra_float_digits = 0;
-- When GROUP BY clause matches with PARTITION KEY.
-- Plan with partitionwise aggregates is disabled
@@ -3240,9 +3261,86 @@ EXPLAIN (VERBOSE, COSTS OFF)
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
SELECT a, count(t1) FROM pagg_tab t1 GROUP BY a HAVING avg(b) < 22 ORDER BY 1;
--- When GROUP BY clause does not match with PARTITION KEY.
-EXPLAIN (COSTS OFF)
+-- Partial aggregates are safe to push down when there is a HAVING clause
+EXPLAIN (VERBOSE, COSTS OFF)
SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b HAVING sum(a) < 700 ORDER BY 1;
+
+-- Partial aggregates are safe to push down without having clause
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, avg(a), max(a), count(*) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+-- Partial aggregates are safe to push down even if we need both variable and variable-based expression
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(a), max(a), count(*), (b/2)::numeric FROM pagg_tab GROUP BY b/2 ORDER BY 4;
+SELECT avg(a), max(a), count(*), (b/2)::numeric FROM pagg_tab GROUP BY b/2 ORDER BY 4;
+
+-- Partial aggregates pushdown behaves sane with non-shipped grouping elements
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT avg(a), (b/2)::numeric, 'test' test FROM pagg_tab GROUP BY 3, b/2 ORDER BY 1, 2 LIMIT 10;
+SELECT avg(a), (b/2)::numeric, 'test' test FROM pagg_tab GROUP BY 3, b/2 ORDER BY 1, 2 LIMIT 10;
+
+-- Partial aggregates are safe to push down for all built-in aggregates
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT
+ /* The cases that don't require import or export functions */
+ bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8),
+ bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8),
+ bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8),
+ bool_and(c_bool),
+ bool_or(c_bool),
+ every(c_bool),
+ max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_1b), max(c_xid8),
+ min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), max(c_1b), min(c_xid8),
+ avg(b::int4)
+ FROM pagg_tab WHERE c_serial between 1 and 30;
+
+SELECT
+ /* The cases that don't require import or export functions */
+ bit_and(c_bit), bit_and(c_1or3int2), bit_and(c_1or3int4), bit_and(c_1or3int8),
+ bit_or(c_bit), bit_or(c_1or3int2), bit_or(c_1or3int4), bit_or(c_1or3int8),
+ bit_xor(c_bit), bit_xor(c_1or3int2), bit_xor(c_1or3int4), bit_xor(c_1or3int8),
+ bool_and(c_bool),
+ bool_or(c_bool),
+ every(c_bool),
+ max(c_1c::char(1)), max('2000-01-01'::date + b), max('0.0.0.0'::inet + b), max(b::float4), max(b::float8), max(b::int2), max(b::int4), max(b::int8), max(c_interval), max(c_money), max(b::numeric), max(b::oid), max(c_pg_lsn), max(c_tid), max(c_1c), max(c_time), max(c_timetz), max(c_timestamp), max(c_timestamptz), max(c_1b), max(c_xid8),
+ min(c_1c::char(1)), min('2000-01-01'::date + b), min('0.0.0.0'::inet + b), min(b::float4), min(b::float8), min(b::int2), min(b::int4), min(b::int8), min(c_interval), min(c_money), min(b::numeric), min(b::oid), min(c_pg_lsn), min(c_tid), min(c_1c), min(c_time), min(c_timetz), min(c_timestamp), min(c_timestamptz), max(c_1b), min(c_xid8),
+ avg(b::int4)
+ FROM pagg_tab WHERE c_serial between 1 and 30;
+
+-- Tests for partial_aggregate_support
+ALTER SERVER loopback OPTIONS (ADD partial_aggregate_support 'false');
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+ALTER SERVER loopback OPTIONS (SET partial_aggregate_support 'true');
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+SELECT b, max(a) FROM pagg_tab GROUP BY b ORDER BY 1;
+
+ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '1.0');
+SET enable_partitionwise_join=on;
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(t1.b) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a);
+SELECT max(t1.b) FROM pagg_tab t1 JOIN pagg_tab t2 USING(a);
+RESET enable_partitionwise_join;
+ALTER SERVER loopback OPTIONS (SET fdw_tuple_cost '0.1');
+
+ALTER SERVER loopback OPTIONS (DROP partial_aggregate_support);
+
+-- It is unsafe to push down partial aggregates which contain DISTINCT clauses
+EXPLAIN (VERBOSE, COSTS OFF)
+SELECT max(a), count(distinct b) FROM pagg_tab;
+SELECT max(a), count(distinct b) FROM pagg_tab;
+
+-- It is unsafe to push down partial aggregates which contain ORDER BY clauses
+EXPLAIN (VERBOSE, COSTS OFF) SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30;
+SELECT array_agg(b order by b) FROM pagg_tab WHERE c_serial between 1 and 30;
+
+ALTER SERVER loopback OPTIONS (DROP fdw_tuple_cost);
+RESET TimeZone;
-- ===================================================================
-- access rights and superuser
diff --git a/doc/src/sgml/postgres-fdw.sgml b/doc/src/sgml/postgres-fdw.sgml
index a7f2f5ca182..c3a6e3b1975 100644
--- a/doc/src/sgml/postgres-fdw.sgml
+++ b/doc/src/sgml/postgres-fdw.sgml
@@ -453,6 +453,44 @@ OPTIONS (ADD password_required 'false');
</sect3>
+ <sect3 id="postgres-fdw-options-partial-aggregate-pushdown">
+ <title>Partial Aggregate Pushdown Option</title>
+
+ <para>
+ When performing partial aggregate pushdown,
+ <filename>postgres_fdw</filename> gets the state value from the local
+ server; see <xref linkend="partial-aggregate-pushdown"/>
+ for partial aggregate pushdown.
+ By default, <filename>postgres_fdw</filename> assumes that the format
+ of state values on the local server is the same as the format of state
+ values on the remote server for every aggregate function. This may be
+ overridden using the following option:
+ </para>
+
+ <variablelist>
+
+ <varlistentry>
+ <term><literal>partial_aggregate_support</literal> (<type>boolean</type>)</term>
+ <listitem>
+ <para>
+ If this option is false, <filename>postgres_fdw</filename>
+ assumes that the format of state values on the local server is
+ the same as the format of state values on the remote server for
+ every aggregate function. If this option is true,
+ during query planning, <filename>postgres_fdw</filename> will
+ connect to the remote server and retrieve the remote server version.
+ If the remote version is the same, the format of state values on the
+ local server will be assumed to the same as the format of state
+ values on the remote server for every aggregate function.
+ If not same, <filename>postgres_fdw</filename> will not perform
+ partial aggregate pushdown. The default is <literal>false</literal>.
+ </para>
+ </listitem>
+ </varlistentry>
+
+ </variablelist>
+ </sect3>
+
<sect3 id="postgres-fdw-options-asynchronous-execution">
<title>Asynchronous Execution Options</title>
@@ -1095,14 +1133,15 @@ postgres=# SELECT postgres_fdw_disconnect_all();
<para>
<filename>postgres_fdw</filename> attempts to optimize remote queries to reduce
the amount of data transferred from foreign servers. This is done by
- sending query <literal>WHERE</literal> clauses to the remote server for
- execution, and by not retrieving table columns that are not needed for
- the current query. To reduce the risk of misexecution of queries,
- <literal>WHERE</literal> clauses are not sent to the remote server unless they use
- only data types, operators, and functions that are built-in or belong to an
- extension that's listed in the foreign server's <literal>extensions</literal>
- option. Operators and functions in such clauses must
- be <literal>IMMUTABLE</literal> as well.
+ sending query <literal>WHERE</literal> clauses and aggregate expressions
+ to the remote server for execution, and by not retrieving table
+ columns that are not needed for the current query. To reduce the
+ risk of misexecution of queries, <literal>WHERE</literal> clauses
+ and aggregate expressions are not sent to the remote server unless
+ they only use data types, operators, and functions that are built-in
+ or belong to an extension that is listed in the foreign server's
+ <literal>extensions</literal> option. Operators and functions in such
+ clauses must be <literal>IMMUTABLE</literal> as well.
For an <command>UPDATE</command> or <command>DELETE</command> query,
<filename>postgres_fdw</filename> attempts to optimize the query execution by
sending the whole query to the remote server if there are no query
@@ -1132,6 +1171,54 @@ postgres=# SELECT postgres_fdw_disconnect_all();
</para>
</sect2>
+ <sect2 id="partial-aggregate-pushdown">
+ <title>Partial Aggregate Pushdown</title>
+ <para>
+ Partial aggregate pushdown is an optimization for queries that contain
+ aggregate expressions for a partitioned table across one or more remote
+ servers. If multiple conditions are met, aggregate expressions with
+ <literal>PARTIAL_AGGREGATE</literal> keyword are sent to the
+ remote servers. The conditions under which this sending is active are
+ as follows.
+ <itemizedlist spacing="compact">
+ <listitem>
+ <para>
+ The format of state values on the local server is the same as the
+ format of state values on the remote server. See <xref
+ linkend="postgres-fdw-options-partial-aggregate-pushdown"/> for the detail
+ </para>
+ </listitem>
+ <listitem>
+ <para>
+ The aggregate expressions in the query do not contain
+ <literal>DISTINCT</literal> or <literal>ORDER BY</literal> clauses
+ </para>
+ </listitem>
+ <listitem>
+ <para>
+ The data type of the state value is neither <type>internal</type>
+ nor pseudo-type.
+ </para>
+ </listitem>
+ </itemizedlist>
+ When performing partial aggregate pushdown,
+ <filename>postgres_fdw</filename> will not send <literal>HAVING</literal>
+ clause. For example, let us assume that there is one remote server
+ and the local server receive the following
+ query, where <literal>f_t</literal> is a foreign table and
+ <literal>c_integer</literal> is a column whose type is <literal>integer</literal>.
+ And let us assume that <literal>f_t</literal> corresponds
+ the table <literal>t</literal> on the local server.
+<programlisting>
+SELECT avg(c_integer) FROM f_t HAVING sum(c_integer) > 0
+</programlisting>
+ Then <filename>postgres_fdw</filename> send the following query to the remote servers.
+<programlisting>
+SELECT avg(c_integer) PARTIAL_AGGREGATE FROM t
+</programlisting>
+ </para>
+ </sect2>
+
<sect2 id="postgres-fdw-remote-query-execution-environment">
<title>Remote Query Execution Environment</title>
diff --git a/doc/src/sgml/syntax.sgml b/doc/src/sgml/syntax.sgml
index 916189a7d68..967ddd3be39 100644
--- a/doc/src/sgml/syntax.sgml
+++ b/doc/src/sgml/syntax.sgml
@@ -1598,10 +1598,10 @@ sqrt(2)
syntax of an aggregate expression is one of the following:
<synopsis>
-<replaceable>aggregate_name</replaceable> (<replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ]
-<replaceable>aggregate_name</replaceable> (ALL <replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ]
-<replaceable>aggregate_name</replaceable> (DISTINCT <replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ]
-<replaceable>aggregate_name</replaceable> ( * ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ]
+<replaceable>aggregate_name</replaceable> (<replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ] [ PARTIAL_AGGREGATE ]
+<replaceable>aggregate_name</replaceable> (ALL <replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ] [ PARTIAL_AGGREGATE ]
+<replaceable>aggregate_name</replaceable> (DISTINCT <replaceable>expression</replaceable> [ , ... ] [ <replaceable>order_by_clause</replaceable> ] ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ] [ PARTIAL_AGGREGATE ]
+<replaceable>aggregate_name</replaceable> ( * ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ] [ PARTIAL_AGGREGATE ]
<replaceable>aggregate_name</replaceable> ( [ <replaceable>expression</replaceable> [ , ... ] ] ) WITHIN GROUP ( <replaceable>order_by_clause</replaceable> ) [ FILTER ( WHERE <replaceable>filter_clause</replaceable> ) ]
</synopsis>
@@ -1779,6 +1779,25 @@ FROM generate_series(1,10) AS s(i);
</programlisting>
</para>
+ <para>
+ If <literal>PARTIAL_AGGREGATE</literal> is specified, then the aggregate
+ expression yields the value before finalization of the aggregation.
+ For example, the data type of the state value of <literal>avg ( integer )</literal> is <literal>bigint[]</literal>.
+ The state value of <literal>avg ( integer )</literal> has two elements, which are the total number
+ of input values and the sum of the input values. Therefore, an example of a
+ call of <literal>avg ( integer )</literal> with <literal>PARTIAL_AGGREGATE</literal> is:
+<programlisting>
+WITH vals (v) AS ( VALUES (1),(5) )
+SELECT
+ avg(v) PARTIAL_AGGREGATE
+FROM vals;
+ avg
+-------
+ {2,3}
+(1 row)
+</programlisting>
+ </para>
+
<para>
The predefined aggregate functions are described in <xref
linkend="functions-aggregate"/>. Other aggregate functions can be added
diff --git a/src/backend/executor/nodeAgg.c b/src/backend/executor/nodeAgg.c
index b4a7698a0b3..4eed281d78b 100644
--- a/src/backend/executor/nodeAgg.c
+++ b/src/backend/executor/nodeAgg.c
@@ -1074,9 +1074,11 @@ finalize_aggregate(AggState *aggstate,
}
/*
- * Apply the agg's finalfn if one is provided, else return transValue.
+ * Apply the agg's finalfn if one is provided and PARTIAL_AGGREGATE
+ * keyword is not specified, else return transValue.
*/
- if (OidIsValid(peragg->finalfn_oid))
+ if (OidIsValid(peragg->finalfn_oid) &&
+ (peragg->aggref->agg_partial == false))
{
int numFinalArgs = peragg->numFinalArgs;
diff --git a/src/backend/nodes/makefuncs.c b/src/backend/nodes/makefuncs.c
index e2d9e9be41a..c17270a9457 100644
--- a/src/backend/nodes/makefuncs.c
+++ b/src/backend/nodes/makefuncs.c
@@ -685,6 +685,7 @@ makeFuncCall(List *name, List *args, CoercionForm funcformat, int location)
n->agg_within_group = false;
n->agg_star = false;
n->agg_distinct = false;
+ n->agg_partial = false;
n->func_variadic = false;
n->funcformat = funcformat;
n->location = location;
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index a4d523dcb0f..9c9bcab648a 100644
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -209,7 +209,7 @@ static PathTarget *make_group_input_target(PlannerInfo *root,
PathTarget *final_target);
static PathTarget *make_partial_grouping_target(PlannerInfo *root,
PathTarget *grouping_target,
- Node *havingQual);
+ GroupPathExtraData *extra);
static List *postprocess_setop_tlist(List *new_tlist, List *orig_tlist);
static void optimize_window_clauses(PlannerInfo *root,
WindowFuncLists *wflists);
@@ -5516,6 +5516,99 @@ make_group_input_target(PlannerInfo *root, PathTarget *final_target)
return set_pathtarget_cost_width(root, input_target);
}
+/*
+ * setGroupClausePartial
+ * Generate a groupClause and a pathtarget for partial aggregate
+ * pushdown by FDW and set them to GroupPathExtraData.
+ */
+static void
+setGroupClausePartial(PathTarget *partial_target, List *non_group_exprs,
+ List *groupClause, GroupPathExtraData *extra)
+{
+ int exprno,
+ refno;
+ ListCell *lc;
+ Index maxRef = 0;
+ List *exprs_processed = NIL;
+ int exprs_num = 0;
+
+ foreach(lc, groupClause)
+ {
+ SortGroupClause *sgc = (SortGroupClause *) lfirst(lc);
+
+ if (sgc->tleSortGroupRef > maxRef)
+ maxRef = sgc->tleSortGroupRef;
+ }
+ maxRef++;
+
+ extra->groupClausePartial = list_copy_deep(groupClause);
+ extra->partial_target = copy_pathtarget(partial_target);
+
+ if (partial_target->exprs)
+ exprs_num = partial_target->exprs->length;
+
+ foreach(lc, non_group_exprs)
+ {
+ Expr *expr = (Expr *) lfirst(lc);
+
+ refno = -1;
+ if (list_member(exprs_processed, expr) ||
+ (!IsA(expr, Var) && !IsA(expr, PlaceHolderVar)))
+ continue;
+ exprs_processed = lappend(exprs_processed, expr);
+ for (exprno = 0; exprno < exprs_num; exprno++)
+ {
+ Expr *target_expr = (Expr *) list_nth(partial_target->exprs, exprno);
+
+ if (equal(target_expr, expr))
+ {
+ refno = exprno;
+ break;
+ }
+ }
+ if (refno < 0)
+ {
+ SortGroupClause *grpcl = makeNode(SortGroupClause);
+
+ grpcl->tleSortGroupRef = maxRef++;
+ extra->groupClausePartial = lappend(extra->groupClausePartial, grpcl);
+ add_column_to_pathtarget(extra->partial_target, expr, grpcl->tleSortGroupRef);
+ }
+ }
+}
+
+/*
+ * adjustAggrefForPartial
+ * Adjust Aggrefs to put them in partial mode
+ */
+static void
+adjustAggrefForPartial(List *exprs)
+{
+ ListCell *lc;
+
+ foreach(lc, exprs)
+ {
+ Aggref *aggref = (Aggref *) lfirst(lc);
+
+ if (IsA(aggref, Aggref))
+ {
+ Aggref *newaggref;
+
+ /*
+ * We shouldn't need to copy the substructure of the Aggref node,
+ * but flat-copy the node itself to avoid damaging other trees.
+ */
+ newaggref = makeNode(Aggref);
+ memcpy(newaggref, aggref, sizeof(Aggref));
+
+ /* For now, assume serialization is required */
+ mark_partial_aggref(newaggref, AGGSPLIT_INITIAL_SERIAL);
+
+ lfirst(lc) = newaggref;
+ }
+ }
+}
+
/*
* make_partial_grouping_target
* Generate appropriate PathTarget for output of partial aggregate
@@ -5531,11 +5624,15 @@ make_group_input_target(PlannerInfo *root, PathTarget *final_target)
*
* grouping_target is the tlist to be emitted by the topmost aggregation step.
* havingQual represents the HAVING clause.
+ *
+ * Modified PathTarget cannot be used by FDW as-is to deparse this statement.
+ * So, before modifying PathTarget, setGroupClausePartial generates
+ * another Pathtarget and another list of SortGroupClauses.
*/
static PathTarget *
make_partial_grouping_target(PlannerInfo *root,
PathTarget *grouping_target,
- Node *havingQual)
+ GroupPathExtraData *extra)
{
PathTarget *partial_target;
List *non_group_cols;
@@ -5577,8 +5674,8 @@ make_partial_grouping_target(PlannerInfo *root,
/*
* If there's a HAVING clause, we'll need the Vars/Aggrefs it uses, too.
*/
- if (havingQual)
- non_group_cols = lappend(non_group_cols, havingQual);
+ if (extra->havingQual)
+ non_group_cols = lappend(non_group_cols, extra->havingQual);
/*
* Pull out all the Vars, PlaceHolderVars, and Aggrefs mentioned in
@@ -5591,35 +5688,17 @@ make_partial_grouping_target(PlannerInfo *root,
PVC_INCLUDE_AGGREGATES |
PVC_RECURSE_WINDOWFUNCS |
PVC_INCLUDE_PLACEHOLDERS);
-
+ setGroupClausePartial(partial_target, non_group_exprs, root->processed_groupClause, extra);
add_new_columns_to_pathtarget(partial_target, non_group_exprs);
+ add_new_columns_to_pathtarget(extra->partial_target, non_group_exprs);
/*
* Adjust Aggrefs to put them in partial mode. At this point all Aggrefs
* are at the top level of the target list, so we can just scan the list
* rather than recursing through the expression trees.
*/
- foreach(lc, partial_target->exprs)
- {
- Aggref *aggref = (Aggref *) lfirst(lc);
-
- if (IsA(aggref, Aggref))
- {
- Aggref *newaggref;
-
- /*
- * We shouldn't need to copy the substructure of the Aggref node,
- * but flat-copy the node itself to avoid damaging other trees.
- */
- newaggref = makeNode(Aggref);
- memcpy(newaggref, aggref, sizeof(Aggref));
-
- /* For now, assume serialization is required */
- mark_partial_aggref(newaggref, AGGSPLIT_INITIAL_SERIAL);
-
- lfirst(lc) = newaggref;
- }
- }
+ adjustAggrefForPartial(partial_target->exprs);
+ adjustAggrefForPartial(extra->partial_target->exprs);
/* clean up cruft */
list_free(non_group_exprs);
@@ -7280,7 +7359,7 @@ create_partial_grouping_paths(PlannerInfo *root,
*/
partially_grouped_rel->reltarget =
make_partial_grouping_target(root, grouped_rel->reltarget,
- extra->havingQual);
+ extra);
if (!extra->partial_costs_set)
{
diff --git a/src/backend/optimizer/plan/setrefs.c b/src/backend/optimizer/plan/setrefs.c
index 999a5a8ab5a..2d6858e125c 100644
--- a/src/backend/optimizer/plan/setrefs.c
+++ b/src/backend/optimizer/plan/setrefs.c
@@ -2651,6 +2651,7 @@ convert_combining_aggrefs(Node *node, void *context)
parent_agg = copyObject(child_agg);
child_agg->args = orig_agg->args;
child_agg->aggfilter = orig_agg->aggfilter;
+ child_agg->agg_partial = orig_agg->agg_partial;
/*
* Now, set up child_agg to represent the first phase of partial
diff --git a/src/backend/optimizer/prep/prepagg.c b/src/backend/optimizer/prep/prepagg.c
index c0a2f04a8c3..c3ffd36d6e4 100644
--- a/src/backend/optimizer/prep/prepagg.c
+++ b/src/backend/optimizer/prep/prepagg.c
@@ -410,6 +410,7 @@ find_compatible_agg(PlannerInfo *root, Aggref *newagg,
/* all of the following must be the same or it's no match */
if (newagg->inputcollid != existingRef->inputcollid ||
newagg->aggtranstype != existingRef->aggtranstype ||
+ newagg->agg_partial != existingRef->agg_partial ||
newagg->aggstar != existingRef->aggstar ||
newagg->aggvariadic != existingRef->aggvariadic ||
newagg->aggkind != existingRef->aggkind ||
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
index 271ae26cbaf..11a49da779f 100644
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -629,6 +629,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
%type <list> within_group_clause
%type <node> filter_clause
+%type <boolean> partial_aggregate_clause
%type <list> window_clause window_definition_list opt_partition_clause
%type <windef> window_definition over_clause window_specification
opt_frame_clause frame_extent frame_bound
@@ -756,7 +757,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
ORDER ORDINALITY OTHERS OUT_P OUTER_P
OVER OVERLAPS OVERLAY OVERRIDING OWNED OWNER
- PARALLEL PARAMETER PARSER PARTIAL PARTITION PASSING PASSWORD PATH
+ PARALLEL PARAMETER PARSER PARTIAL PARTIAL_AGGREGATE PARTITION PASSING PASSWORD PATH
PERIOD PLACING PLAN PLANS POLICY
POSITION PRECEDING PRECISION PRESERVE PREPARE PREPARED PRIMARY
PRIOR PRIVILEGES PROCEDURAL PROCEDURE PROCEDURES PROGRAM PUBLICATION
@@ -15758,10 +15759,11 @@ func_application: func_name '(' ')'
* (Note that many of the special SQL functions wouldn't actually make any
* sense as functional index entries, but we ignore that consideration here.)
*/
-func_expr: func_application within_group_clause filter_clause over_clause
+func_expr: func_application within_group_clause filter_clause partial_aggregate_clause over_clause
{
FuncCall *n = (FuncCall *) $1;
+ n->agg_partial = $4;
/*
* The order clause for WITHIN GROUP and the one for
* plain-aggregate ORDER BY share a field, so we have to
@@ -15782,6 +15784,11 @@ func_expr: func_application within_group_clause filter_clause over_clause
(errcode(ERRCODE_SYNTAX_ERROR),
errmsg("cannot use DISTINCT with WITHIN GROUP"),
parser_errposition(@2)));
+ if (n->agg_partial)
+ ereport(ERROR,
+ (errcode(ERRCODE_SYNTAX_ERROR),
+ errmsg("cannot use PARTIAL_AGGREGATE with WITHIN GROUP"),
+ parser_errposition(@2)));
if (n->func_variadic)
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -15791,7 +15798,7 @@ func_expr: func_application within_group_clause filter_clause over_clause
n->agg_within_group = true;
}
n->agg_filter = $3;
- n->over = $4;
+ n->over = $5;
$$ = (Node *) n;
}
| json_aggregate_func filter_clause over_clause
@@ -16383,6 +16390,10 @@ filter_clause:
| /*EMPTY*/ { $$ = NULL; }
;
+partial_aggregate_clause:
+ PARTIAL_AGGREGATE { $$ = true; }
+ | /*EMPTY*/ { $$ = false; }
+ ;
/*
* Window Definitions
@@ -17905,6 +17916,7 @@ unreserved_keyword:
| PARAMETER
| PARSER
| PARTIAL
+ | PARTIAL_AGGREGATE
| PARTITION
| PASSING
| PASSWORD
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
index 0ac8966e30f..cd23eeae248 100644
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -16,6 +16,7 @@
#include "access/htup_details.h"
#include "catalog/pg_aggregate.h"
+#include "catalog/pg_proc.h"
#include "catalog/pg_constraint.h"
#include "catalog/pg_type.h"
#include "common/int.h"
@@ -106,8 +107,8 @@ static Node *make_agg_arg(Oid argtype, Oid argcollation);
* pstate level.
*/
void
-transformAggregateCall(ParseState *pstate, Aggref *agg,
- List *args, List *aggorder, bool agg_distinct)
+transformAggregateCall(ParseState *pstate, Aggref *agg, List *args,
+ List *aggorder, bool agg_distinct, bool agg_partial)
{
List *argtypes = NIL;
List *tlist = NIL;
@@ -152,8 +153,8 @@ transformAggregateCall(ParseState *pstate, Aggref *agg,
torder, tlist, sortby);
}
- /* Never any DISTINCT in an ordered-set agg */
- Assert(!agg_distinct);
+ /* Never any DISTINCT and PARTIAL_AGG in an ordered-set agg */
+ Assert(!agg_distinct || !agg_partial);
}
else
{
@@ -227,6 +228,18 @@ transformAggregateCall(ParseState *pstate, Aggref *agg,
agg->args = tlist;
agg->aggorder = torder;
agg->aggdistinct = tdistinct;
+ agg->agg_partial = agg_partial;
+ if(agg->agg_partial){
+ HeapTuple aggtup;
+ Form_pg_aggregate aggform;
+
+ aggtup = SearchSysCache1(AGGFNOID, ObjectIdGetDatum(agg->aggfnoid));
+ if (!HeapTupleIsValid(aggtup))
+ elog(ERROR, "cache lookup failed for aggregate %u", agg->aggfnoid);
+ aggform = (Form_pg_aggregate) GETSTRUCT(aggtup);
+ ReleaseSysCache(aggtup);
+ agg->aggtype = aggform->aggtranstype;
+ }
/*
* Now build the aggargtypes list with the type OIDs of the direct and
diff --git a/src/backend/parser/parse_clause.c b/src/backend/parser/parse_clause.c
index 2e64fcae7b2..4351846c5a8 100644
--- a/src/backend/parser/parse_clause.c
+++ b/src/backend/parser/parse_clause.c
@@ -539,6 +539,7 @@ transformRangeFunction(ParseState *pstate, RangeFunction *r)
fc->over == NULL &&
!fc->agg_star &&
!fc->agg_distinct &&
+ !fc->agg_partial &&
!fc->func_variadic &&
coldeflist == NIL)
{
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
index 9caf1e481a2..cac8ba8c080 100644
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -3893,7 +3893,8 @@ transformJsonAggConstructor(ParseState *pstate, JsonAggConstructor *agg_ctor,
aggref->aggtransno = -1;
aggref->location = agg_ctor->location;
- transformAggregateCall(pstate, aggref, args, agg_ctor->agg_order, false);
+ transformAggregateCall(pstate, aggref, args, agg_ctor->agg_order, false,
+ false);
node = (Node *) aggref;
}
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
index 583bbbf232f..463b1dfdb33 100644
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -97,6 +97,7 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
bool agg_within_group = (fn ? fn->agg_within_group : false);
bool agg_star = (fn ? fn->agg_star : false);
bool agg_distinct = (fn ? fn->agg_distinct : false);
+ bool agg_partial = (fn ? fn->agg_partial : false);
bool func_variadic = (fn ? fn->func_variadic : false);
CoercionForm funcformat = (fn ? fn->funcformat : COERCE_EXPLICIT_CALL);
bool could_be_projection;
@@ -222,7 +223,8 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
*/
could_be_projection = (nargs == 1 && !proc_call &&
agg_order == NIL && agg_filter == NULL &&
- !agg_star && !agg_distinct && over == NULL &&
+ !agg_star && !agg_distinct &&
+ !agg_partial && over == NULL &&
!func_variadic && argnames == NIL &&
list_length(funcname) == 1 &&
(actual_arg_types[0] == RECORDOID ||
@@ -322,6 +324,12 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
errmsg("DISTINCT specified, but %s is not an aggregate function",
NameListToString(funcname)),
parser_errposition(pstate, location)));
+ if (agg_partial)
+ ereport(ERROR,
+ (errcode(ERRCODE_WRONG_OBJECT_TYPE),
+ errmsg("PARTIAL_AGGREGATE specified, but %s is not an aggregate function",
+ NameListToString(funcname)),
+ parser_errposition(pstate, location)));
if (agg_within_group)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
@@ -392,6 +400,8 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
parser_errposition(pstate, location)));
/* gram.y rejects DISTINCT + WITHIN GROUP */
Assert(!agg_distinct);
+ /* gram.y rejects PARTIAL_AGGREGATE */
+ Assert(!agg_partial);
/* gram.y rejects VARIADIC + WITHIN GROUP */
Assert(!func_variadic);
@@ -814,7 +824,8 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
parser_errposition(pstate, location)));
/* parse_agg.c does additional aggregate-specific processing */
- transformAggregateCall(pstate, aggref, fargs, agg_order, agg_distinct);
+ transformAggregateCall(pstate, aggref, fargs, agg_order, agg_distinct,
+ agg_partial);
retval = (Node *) aggref;
}
@@ -846,6 +857,15 @@ ParseFuncOrColumn(ParseState *pstate, List *funcname, List *fargs,
errmsg("DISTINCT is not implemented for window functions"),
parser_errposition(pstate, location)));
+ /*
+ * partial aggregates not allowed in windows yet
+ */
+ if (agg_partial)
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("PARTIAL_AGGREGATE is not implemented for window functions"),
+ parser_errposition(pstate, location)));
+
/*
* Reject attempt to call a parameterless aggregate without (*)
* syntax. This is mere pedantry but some folks insisted ...
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
index 9e90acedb91..0d960eab8d2 100644
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -10915,7 +10915,8 @@ get_agg_expr_helper(Aggref *aggref, deparse_context *context,
/* Print the aggregate name, schema-qualified if needed */
appendStringInfo(buf, "%s(%s", funcname,
- (aggref->aggdistinct != NIL) ? "DISTINCT " : "");
+ (aggref->aggdistinct != NIL) ? "DISTINCT " :
+ aggref->agg_partial ? "PARTIAL_AGGREGATE " : "");
if (AGGKIND_IS_ORDERED_SET(aggref->aggkind))
{
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
index 23c9e3c5abf..5af495d2b90 100644
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -442,6 +442,7 @@ typedef struct FuncCall
bool agg_within_group; /* ORDER BY appeared in WITHIN GROUP */
bool agg_star; /* argument was really '*' */
bool agg_distinct; /* arguments were labeled DISTINCT */
+ bool agg_partial; /* arguments were labeled PARTIAL_AGGREGATE */
bool func_variadic; /* last argument was labeled VARIADIC */
CoercionForm funcformat; /* how to display this node */
ParseLoc location; /* token location, or -1 if unknown */
diff --git a/src/include/nodes/pathnodes.h b/src/include/nodes/pathnodes.h
index fbf05322c75..03d4cdec740 100644
--- a/src/include/nodes/pathnodes.h
+++ b/src/include/nodes/pathnodes.h
@@ -3319,6 +3319,9 @@ typedef enum
* havingQual gives list of quals to be applied after aggregation.
* targetList gives list of columns to be projected.
* patype is the type of partitionwise aggregation that is being performed.
+ * groupClausePartial is List of SortGroupClauses for partial aggregate
+ * pushdown by FDW
+ * partial_target is PathTarget for partial aggregate pushdown by FDW
*/
typedef struct
{
@@ -3333,6 +3336,8 @@ typedef struct
Node *havingQual;
List *targetList;
PartitionwiseAggregateType patype;
+ List *groupClausePartial;
+ PathTarget *partial_target;
} GroupPathExtraData;
/*
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
index d0576da3e25..79cfd69d214 100644
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -490,6 +490,9 @@ typedef struct Aggref
/* DISTINCT (list of SortGroupClause) */
List *aggdistinct;
+ /* true if there is PARTIAL_AGGREGATE keyword */
+ bool agg_partial;
+
/* FILTER expression, if any */
Expr *aggfilter;
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
index 40cf090ce61..6d7044078ca 100644
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -335,6 +335,7 @@ PG_KEYWORD("parallel", PARALLEL, UNRESERVED_KEYWORD, BARE_LABEL)
PG_KEYWORD("parameter", PARAMETER, UNRESERVED_KEYWORD, BARE_LABEL)
PG_KEYWORD("parser", PARSER, UNRESERVED_KEYWORD, BARE_LABEL)
PG_KEYWORD("partial", PARTIAL, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("partial_aggregate", PARTIAL_AGGREGATE, UNRESERVED_KEYWORD, AS_LABEL)
PG_KEYWORD("partition", PARTITION, UNRESERVED_KEYWORD, BARE_LABEL)
PG_KEYWORD("passing", PASSING, UNRESERVED_KEYWORD, BARE_LABEL)
PG_KEYWORD("password", PASSWORD, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_agg.h b/src/include/parser/parse_agg.h
index 277d45baf83..e4ca70787c1 100644
--- a/src/include/parser/parse_agg.h
+++ b/src/include/parser/parse_agg.h
@@ -17,7 +17,7 @@
extern void transformAggregateCall(ParseState *pstate, Aggref *agg,
List *args, List *aggorder,
- bool agg_distinct);
+ bool agg_distinct, bool agg_partial);
extern Node *transformGroupingFunc(ParseState *pstate, GroupingFunc *p);
diff --git a/src/test/regress/expected/aggregates.out b/src/test/regress/expected/aggregates.out
index 6b6371c3e74..97c9fb7bcc2 100644
--- a/src/test/regress/expected/aggregates.out
+++ b/src/test/regress/expected/aggregates.out
@@ -3536,6 +3536,14 @@ set work_mem to default;
----+----+----
(0 rows)
+-- PARTIAL_AGGREGATE tests
+-- Check return type of partial aggregate
+select pg_typeof(avg(a::int4) partial_aggregate), pg_typeof(avg(a::int4)) from aggtest;
+ pg_typeof | pg_typeof
+-----------+-----------
+ bigint[] | numeric
+(1 row)
+
drop table agg_group_1;
drop table agg_group_2;
drop table agg_group_3;
diff --git a/src/test/regress/expected/partition_aggregate.out b/src/test/regress/expected/partition_aggregate.out
index 5f2c0cf5786..cfde38c6ed1 100644
--- a/src/test/regress/expected/partition_aggregate.out
+++ b/src/test/regress/expected/partition_aggregate.out
@@ -399,6 +399,35 @@ SELECT a, sum(b order by a) FROM pagg_tab GROUP BY a ORDER BY 1, 2;
-> Seq Scan on pagg_tab_p3 pagg_tab_3
(10 rows)
+-- PARTIAL_AGGREGATE tests
+-- Check partial aggregate over partitioned table
+explain (verbose, costs off)
+select avg(a) partial_aggregate, avg(a) from pagg_tab;
+ QUERY PLAN
+----------------------------------------------------------------------------------------------
+ Finalize Aggregate
+ Output: avg(PARTIAL_AGGREGATE pagg_tab.a), avg(pagg_tab.a)
+ -> Append
+ -> Partial Aggregate
+ Output: PARTIAL avg(PARTIAL_AGGREGATE pagg_tab.a), PARTIAL avg(pagg_tab.a)
+ -> Seq Scan on public.pagg_tab_p1 pagg_tab
+ Output: pagg_tab.a
+ -> Partial Aggregate
+ Output: PARTIAL avg(PARTIAL_AGGREGATE pagg_tab_1.a), PARTIAL avg(pagg_tab_1.a)
+ -> Seq Scan on public.pagg_tab_p2 pagg_tab_1
+ Output: pagg_tab_1.a
+ -> Partial Aggregate
+ Output: PARTIAL avg(PARTIAL_AGGREGATE pagg_tab_2.a), PARTIAL avg(pagg_tab_2.a)
+ -> Seq Scan on public.pagg_tab_p3 pagg_tab_2
+ Output: pagg_tab_2.a
+(15 rows)
+
+select avg(a) partial_aggregate, avg(a) from pagg_tab;
+ avg | avg
+--------------+--------------------
+ {3000,28500} | 9.5000000000000000
+(1 row)
+
-- JOIN query
CREATE TABLE pagg_tab1(x int, y int) PARTITION BY RANGE(x);
CREATE TABLE pagg_tab1_p1 PARTITION OF pagg_tab1 FOR VALUES FROM (0) TO (10);
diff --git a/src/test/regress/sql/aggregates.sql b/src/test/regress/sql/aggregates.sql
index 2c47a462b7e..2b2d4640798 100644
--- a/src/test/regress/sql/aggregates.sql
+++ b/src/test/regress/sql/aggregates.sql
@@ -1617,6 +1617,10 @@ set work_mem to default;
union all
(select * from agg_group_4 except select * from agg_hash_4);
+-- PARTIAL_AGGREGATE tests
+-- Check return type of partial aggregate
+select pg_typeof(avg(a::int4) partial_aggregate), pg_typeof(avg(a::int4)) from aggtest;
+
drop table agg_group_1;
drop table agg_group_2;
drop table agg_group_3;
diff --git a/src/test/regress/sql/partition_aggregate.sql b/src/test/regress/sql/partition_aggregate.sql
index ab070fee244..60356a89f1c 100644
--- a/src/test/regress/sql/partition_aggregate.sql
+++ b/src/test/regress/sql/partition_aggregate.sql
@@ -92,6 +92,11 @@ SELECT c, sum(b order by a) FROM pagg_tab GROUP BY c ORDER BY 1, 2;
EXPLAIN (COSTS OFF)
SELECT a, sum(b order by a) FROM pagg_tab GROUP BY a ORDER BY 1, 2;
+-- PARTIAL_AGGREGATE tests
+-- Check partial aggregate over partitioned table
+explain (verbose, costs off)
+select avg(a) partial_aggregate, avg(a) from pagg_tab;
+select avg(a) partial_aggregate, avg(a) from pagg_tab;
-- JOIN query
--
2.39.3
^ permalink raw reply [nested|flat] 42+ messages in thread
* Re: Partial aggregates pushdown
2025-03-28 02:00 RE: Partial aggregates pushdown [email protected] <[email protected]>
@ 2025-03-29 22:22 ` Bruce Momjian <[email protected]>
0 siblings, 0 replies; 42+ messages in thread
From: Bruce Momjian @ 2025-03-29 22:22 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: Tomas Vondra <[email protected]>; Jelte Fennema-Nio <[email protected]>; pgsql-hackers; Robert Haas <[email protected]>; Tom Lane <[email protected]>; Stephen Frost <[email protected]>; Andres Freund <[email protected]>; Tomas Vondra <[email protected]>; Julien Rouhaud <[email protected]>; Daniel Gustafsson <[email protected]>; vignesh C <[email protected]>; Alexander Pyhalov <[email protected]>
On Fri, Mar 28, 2025 at 02:00:44AM +0000, [email protected] wrote:
> Hi Bruce, Jelte, hackers.
>
> I apologize for my late response.
>
> > From: Jelte Fennema-Nio <[email protected]>
> > Sent: Thursday, August 8, 2024 8:49 PM
> > SUMMARY OF THREAD
> >
> > The design of patch 0001 is agreed upon by everyone on the thread (so far). This adds the PARTIAL_AGGREGATE label for
> > aggregates, which will cause the finalfunc not to run. It also starts using PARTIAL_AGGREGATE for pushdown of
> > aggregates in postgres_fdw. In 0001 PARTIAL_AGGREGATE is only supported for aggregates with a non-internal/pseudo
> > type as the stype.
> >
> > The design for patch 0002 is still under debate. This would expand on the functionality added by adding support for
> > PARTIAL_AGGREGATE for aggregates with an internal stype. This is done by returning a byte array containing the bytes
> > that the serialfunc of the aggregate returns.
> >
> > A competing proposal for 0002 is to instead change aggregates to not use an internal stype anymore, and create dedicated
> > types. The main downside here is that infunc and outfunc would need to be added for text serialization, in addition to the
> > binary serialization. An open question is: Can we change the requirements for CREATE TYPE, so that types can be created
> > without infunc and outfunc.
>
> I rebased the patch 0001 and add the documentation to it.
Okay, this is too late for PG 18 but I am hopeful we can make progress
on this for PG 19.
--
Bruce Momjian <[email protected]> https://momjian.us
EDB https://enterprisedb.com
Do not let urgent matters crowd out time for investment in the future.
^ permalink raw reply [nested|flat] 42+ messages in thread
end of thread, other threads:[~2025-03-29 22:22 UTC | newest]
Thread overview: 42+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2021-10-15 13:15 Partial aggregates pushdown Alexander Pyhalov <[email protected]>
2021-10-15 14:56 ` Tomas Vondra <[email protected]>
2021-10-15 15:05 ` Alexander Pyhalov <[email protected]>
2021-10-15 15:26 ` Tomas Vondra <[email protected]>
2021-10-15 19:31 ` Stephen Frost <[email protected]>
2021-10-15 21:59 ` Tomas Vondra <[email protected]>
2021-10-19 06:56 ` Alexander Pyhalov <[email protected]>
2021-10-19 13:25 ` Tomas Vondra <[email protected]>
2021-10-21 10:55 ` Alexander Pyhalov <[email protected]>
2021-11-01 09:47 ` Peter Eisentraut <[email protected]>
2021-11-01 10:30 ` Alexander Pyhalov <[email protected]>
2021-11-01 21:53 ` Ilya Gladyshev <[email protected]>
2021-11-01 23:49 ` Tomas Vondra <[email protected]>
2021-11-01 21:31 ` Ilya Gladyshev <[email protected]>
2021-11-01 21:57 ` Tomas Vondra <[email protected]>
2021-11-02 09:12 ` Alexander Pyhalov <[email protected]>
2021-11-03 13:45 ` Daniel Gustafsson <[email protected]>
2021-11-03 14:50 ` Alexander Pyhalov <[email protected]>
2021-11-15 10:16 ` Daniel Gustafsson <[email protected]>
2021-11-15 13:01 ` Alexander Pyhalov <[email protected]>
2022-08-01 05:55 ` [email protected] <[email protected]>
2022-11-22 01:01 ` [email protected] <[email protected]>
2022-11-22 05:00 ` Ted Yu <[email protected]>
2022-11-22 09:11 ` [email protected] <[email protected]>
2022-11-22 10:51 ` Ted Yu <[email protected]>
2022-11-22 16:05 ` Alexander Pyhalov <[email protected]>
2022-11-30 03:10 ` [email protected] <[email protected]>
2022-11-30 08:12 ` Alexander Pyhalov <[email protected]>
2022-11-30 10:01 ` [email protected] <[email protected]>
2022-11-30 13:30 ` Alexander Pyhalov <[email protected]>
2022-11-30 14:30 ` Alexander Pyhalov <[email protected]>
2022-12-01 02:23 ` [email protected] <[email protected]>
2022-12-01 16:36 ` Alexander Pyhalov <[email protected]>
2022-12-05 02:03 ` [email protected] <[email protected]>
2022-12-07 18:59 ` Andres Freund <[email protected]>
2022-12-15 22:23 ` [email protected] <[email protected]>
2021-10-21 21:43 Re: Partial aggregates pushdown Zhihong Yu <[email protected]>
2021-10-22 06:26 ` Alexander Pyhalov <[email protected]>
2023-05-05 18:40 [PATCH] add ruleutils.c support for MERGE Alvaro Herrera <[email protected]>
2024-01-27 01:57 Re: [CAUTION!! freemail] Re: Partial aggregates pushdown vignesh C <[email protected]>
2025-03-28 02:00 RE: Partial aggregates pushdown [email protected] <[email protected]>
2025-03-29 22:22 ` Bruce Momjian <[email protected]>
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox