Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit48d9d8e

Browse files
committed
Fix a couple of planner bugs introduced by the new ability to discard
ORDER BY <constant> as redundant. One is that this means query_planner()has to canonicalize pathkeys even when the query jointree is empty;the canonicalization was always a no-op in such cases before, but no more.Also, we have to guard against thinking that a set-returning function is"constant" for this purpose. Add a couple of regression tests for theseevidently under-tested cases. Per report from Greg Stark and subsequentexperimentation.
1 parentd5eaa63 commit48d9d8e

File tree

4 files changed

+70
-8
lines changed

4 files changed

+70
-8
lines changed

‎src/backend/optimizer/path/equivclass.c‎

Lines changed: 9 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@
1010
* Portions Copyright (c) 1994, Regents of the University of California
1111
*
1212
* IDENTIFICATION
13-
* $PostgreSQL: pgsql/src/backend/optimizer/path/equivclass.c,v 1.2 2007/01/22 20:00:39 tgl Exp $
13+
* $PostgreSQL: pgsql/src/backend/optimizer/path/equivclass.c,v 1.3 2007/07/07 20:46:45 tgl Exp $
1414
*
1515
*-------------------------------------------------------------------------
1616
*/
@@ -328,8 +328,8 @@ add_eq_member(EquivalenceClass *ec, Expr *expr, Relids relids,
328328
/*
329329
* No Vars, assume it's a pseudoconstant. This is correct for
330330
* entries generated from process_equivalence(), because a WHERE
331-
* clause can't contain aggregates and non-volatility was checked
332-
* before process_equivalence() ever got called. But
331+
* clause can't contain aggregatesor SRFs,and non-volatility was
332+
*checkedbefore process_equivalence() ever got called. But
333333
* get_eclass_for_sort_expr() has to work harder. We put the tests
334334
* there not here to save cycles in the equivalence case.
335335
*/
@@ -428,13 +428,15 @@ get_eclass_for_sort_expr(PlannerInfo *root,
428428
false,expr_datatype);
429429

430430
/*
431-
* add_eq_member doesn't check for volatile functions or aggregates,
432-
* but such could appear in sort expressions, so we have to check
433-
* whether its const-marking was correct.
431+
* add_eq_member doesn't check for volatile functions, set-returning
432+
*functions, or aggregates,but such could appear in sort expressions;
433+
*so we have to checkwhether its const-marking was correct.
434434
*/
435435
if (newec->ec_has_const)
436436
{
437-
if (newec->ec_has_volatile||contain_agg_clause((Node*)expr))
437+
if (newec->ec_has_volatile||
438+
expression_returns_set((Node*)expr)||
439+
contain_agg_clause((Node*)expr))
438440
{
439441
newec->ec_has_const= false;
440442
newem->em_is_const= false;

‎src/backend/optimizer/plan/planmain.c‎

Lines changed: 13 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,7 @@
1414
*
1515
*
1616
* IDENTIFICATION
17-
* $PostgreSQL: pgsql/src/backend/optimizer/plan/planmain.c,v 1.101 2007/05/04 01:13:44 tgl Exp $
17+
* $PostgreSQL: pgsql/src/backend/optimizer/plan/planmain.c,v 1.102 2007/07/07 20:46:45 tgl Exp $
1818
*
1919
*-------------------------------------------------------------------------
2020
*/
@@ -106,9 +106,21 @@ query_planner(PlannerInfo *root, List *tlist,
106106
*/
107107
if (parse->jointree->fromlist==NIL)
108108
{
109+
/* We need a trivial path result */
109110
*cheapest_path= (Path*)
110111
create_result_path((List*)parse->jointree->quals);
111112
*sorted_path=NULL;
113+
/*
114+
* We still are required to canonicalize any pathkeys, in case
115+
* it's something like "SELECT 2+2 ORDER BY 1".
116+
*/
117+
root->canon_pathkeys=NIL;
118+
root->query_pathkeys=canonicalize_pathkeys(root,
119+
root->query_pathkeys);
120+
root->group_pathkeys=canonicalize_pathkeys(root,
121+
root->group_pathkeys);
122+
root->sort_pathkeys=canonicalize_pathkeys(root,
123+
root->sort_pathkeys);
112124
return;
113125
}
114126

‎src/test/regress/expected/select.out‎

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -736,3 +736,35 @@ SELECT * FROM foo ORDER BY f1 DESC NULLS LAST;
736736

737737
(7 rows)
738738

739+
--
740+
-- Test some corner cases that have been known to confuse the planner
741+
--
742+
-- ORDER BY on a constant doesn't really need any sorting
743+
SELECT 1 AS x ORDER BY x;
744+
x
745+
---
746+
1
747+
(1 row)
748+
749+
-- But ORDER BY on a set-valued expression does
750+
create function sillysrf(int) returns setof int as
751+
'values (1),(10),(2),($1)' language sql immutable;
752+
select sillysrf(42);
753+
sillysrf
754+
----------
755+
1
756+
10
757+
2
758+
42
759+
(4 rows)
760+
761+
select sillysrf(-1) order by 1;
762+
sillysrf
763+
----------
764+
-1
765+
1
766+
2
767+
10
768+
(4 rows)
769+
770+
drop function sillysrf(int);

‎src/test/regress/sql/select.sql‎

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -186,3 +186,19 @@ SELECT * FROM foo ORDER BY f1;
186186
SELECT*FROM fooORDER BY f1 NULLS FIRST;
187187
SELECT*FROM fooORDER BY f1DESC;
188188
SELECT*FROM fooORDER BY f1DESC NULLS LAST;
189+
190+
--
191+
-- Test some corner cases that have been known to confuse the planner
192+
--
193+
194+
-- ORDER BY on a constant doesn't really need any sorting
195+
SELECT1AS xORDER BY x;
196+
197+
-- But ORDER BY on a set-valued expression does
198+
createfunctionsillysrf(int) returns setofintas
199+
'values (1),(10),(2),($1)' language sql immutable;
200+
201+
select sillysrf(42);
202+
select sillysrf(-1)order by1;
203+
204+
dropfunction sillysrf(int);

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp