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

Commitb200180

Browse files
committed
Repair commits317aba7 et al for -DWRITE_READ_PARSE_PLAN_TREES.
Letting the rewriter keep RangeTblEntry.relid when expanding a viewRTE, without making the outfuncs/readfuncs changes that went alongwith that originally, is more problematic than I realized. It causesWRITE_READ_PARSE_PLAN_TREES testing to fail because outfuncs/readfuncsdon't think relid need be saved in an RTE_SUBQUERY RTE.There doesn't seem to be any other good route to fixing the whole-rowVar problem solved atf4e7756, so we just have to deal with theconsequences. We can make the eventually-produced plan tree safefor WRITE_READ_PARSE_PLAN_TREES by clearing the relid field at theend of planning, as was already being done for the functions field.(The functions field is not problematic here because our abuse of itis strictly local to the planner.) However, there is no nice fix forthe post-rewrite WRITE_READ_PARSE_PLAN_TREES test.The solution adopted here is to remove the post-rewrite test in theaffected branches. That's surely less than ideal, but a couple ofarguments can be made why it's not unacceptable. First, the behaviorof outfuncs/readfuncs for parsetrees in these branches is frozen nomatter what, because of catalog stability requirements. So we're nottesting anything that is going to change. Second, testingWRITE_READ_PARSE_PLAN_TREES at this particular time doesn't correspondto any direct system functionality requirement, neither rule storagenor plan transmission.Reported-by: Andres Freund <andres@anarazel.de>Author: Tom Lane <tgl@sss.pgh.pa.us>Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>Discussion:https://postgr.es/m/3518c50a-ab18-482f-b916-a37263622501@deepbluecap.comBackpatch-through: 13-15
1 parent39af32f commitb200180

File tree

4 files changed

+17
-43
lines changed

4 files changed

+17
-43
lines changed

‎src/backend/optimizer/plan/setrefs.c

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -438,6 +438,15 @@ add_rte_to_flat_rtable(PlannerGlobal *glob, RangeTblEntry *rte)
438438
newrte->colcollations=NIL;
439439
newrte->securityQuals=NIL;
440440

441+
/*
442+
* Also, if it's a subquery RTE, lose the relid that may have been kept to
443+
* signal that it had been a view. We don't want that to escape the
444+
* planner, mainly because doing so breaks -DWRITE_READ_PARSE_PLAN_TREES
445+
* testing thanks to outfuncs/readfuncs not preserving it.
446+
*/
447+
if (newrte->rtekind==RTE_SUBQUERY)
448+
newrte->relid=InvalidOid;
449+
441450
glob->finalrtable=lappend(glob->finalrtable,newrte);
442451

443452
/*

‎src/backend/rewrite/rewriteHandler.c

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1828,8 +1828,8 @@ ApplyRetrieveRule(Query *parsetree,
18281828

18291829
/*
18301830
* Clear fields that should not be set in a subquery RTE. However, we
1831-
* retain the relid to support correct operation of makeWholeRowVar during
1832-
* planning.
1831+
* retain the relidfor now,to support correct operation of
1832+
*makeWholeRowVar duringplanning.
18331833
*/
18341834
rte->relkind=0;
18351835
rte->rellockmode=0;

‎src/backend/tcop/postgres.c

Lines changed: 4 additions & 39 deletions
Original file line numberDiff line numberDiff line change
@@ -803,45 +803,10 @@ pg_rewrite_query(Query *query)
803803
}
804804
#endif
805805

806-
#ifdefWRITE_READ_PARSE_PLAN_TREES
807-
/* Optional debugging check: pass querytree through outfuncs/readfuncs */
808-
{
809-
List*new_list=NIL;
810-
ListCell*lc;
811-
812-
/*
813-
* We currently lack outfuncs/readfuncs support for most utility
814-
* statement types, so only attempt to write/read non-utility queries.
815-
*/
816-
foreach(lc,querytree_list)
817-
{
818-
Query*query=castNode(Query,lfirst(lc));
819-
820-
if (query->commandType!=CMD_UTILITY)
821-
{
822-
char*str=nodeToString(query);
823-
Query*new_query=stringToNodeWithLocations(str);
824-
825-
/*
826-
* queryId is not saved in stored rules, but we must preserve
827-
* it here to avoid breaking pg_stat_statements.
828-
*/
829-
new_query->queryId=query->queryId;
830-
831-
new_list=lappend(new_list,new_query);
832-
pfree(str);
833-
}
834-
else
835-
new_list=lappend(new_list,query);
836-
}
837-
838-
/* This checks both outfuncs/readfuncs and the equal() routines... */
839-
if (!equal(new_list,querytree_list))
840-
elog(WARNING,"outfuncs/readfuncs failed to produce equal parse tree");
841-
else
842-
querytree_list=new_list;
843-
}
844-
#endif
806+
/*
807+
* We don't apply WRITE_READ_PARSE_PLAN_TREES to rewritten query trees,
808+
* because it breaks the hack of preserving relid for rewritten views.
809+
*/
845810

846811
if (Debug_print_rewritten)
847812
elog_node_display(LOG,"rewritten parse tree",querytree_list,

‎src/include/nodes/parsenodes.h

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -985,8 +985,8 @@ typedef struct RangeTblEntry
985985
* As a special case, relid can also be set in RTE_SUBQUERY RTEs. This
986986
* happens when an RTE_RELATION RTE for a view is transformed to an
987987
* RTE_SUBQUERY during rewriting. We keep the relid because it is useful
988-
* during planning, cf makeWholeRowVar. (Itcannotberelied onduring
989-
*execution, because it will not propagate to parallel workers.)
988+
* during planning, cf makeWholeRowVar. (Itwill notbepassed onto the
989+
*executor, however.)
990990
*
991991
* As a special case, RTE_NAMEDTUPLESTORE can also set relid to indicate
992992
* that the tuple format of the tuplestore is the same as the referenced

0 commit comments

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