From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Fri, 27 Feb 2009 23:30:29 +0000 (+0000)
Subject: Temporarily (I hope) disable flattening of IN/EXISTS sublinks that are within
X-Git-Tag: REL9_0_0~2363
X-Git-Url: http://git.osdn.net/view?a=commitdiff_plain;h=07b9936a0f10d746e5076239813a5e938f2f16be;p=pg-rex%2Fsyncrep.git

Temporarily (I hope) disable flattening of IN/EXISTS sublinks that are within
the ON clause of an outer join.  Doing so is semantically correct but results
in de-optimizing queries that were structured to take advantage of the sublink
style of execution, as seen in recent complaint from Kevin Grittner.  Since
the user can get the other behavior by reorganizing his query, having the
flattening happen automatically is just a convenience, and that doesn't
justify breaking existing applications.  Eventually it would be nice to
re-enable this, but that seems to require a significantly different approach
to outer joins in the executor.
---

diff --git a/src/backend/optimizer/prep/prepjointree.c b/src/backend/optimizer/prep/prepjointree.c
index 600facb578..9df13765f4 100644
--- a/src/backend/optimizer/prep/prepjointree.c
+++ b/src/backend/optimizer/prep/prepjointree.c
@@ -16,7 +16,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepjointree.c,v 1.63 2009/02/25 03:30:37 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepjointree.c,v 1.64 2009/02/27 23:30:29 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -220,6 +220,15 @@ pull_up_sublinks_jointree_recurse(PlannerInfo *root, Node *jtnode,
 		 * The point of the available_rels machinations is to ensure that we
 		 * only pull up quals for which that's okay.
 		 *
+		 * XXX for the moment, we refrain from pulling up IN/EXISTS clauses
+		 * appearing in LEFT or RIGHT join conditions.  Although it is
+		 * semantically valid to do so under the above conditions, we end up
+		 * with a query in which the semijoin or antijoin must be evaluated
+		 * below the outer join, which could perform far worse than leaving
+		 * it as a sublink that is executed only for row pairs that meet the
+		 * other join conditions.  Fixing this seems to require considerable
+		 * restructuring of the executor, but maybe someday it can happen.
+		 *
 		 * We don't expect to see any pre-existing JOIN_SEMI or JOIN_ANTI
 		 * nodes here.
 		 */
@@ -232,17 +241,21 @@ pull_up_sublinks_jointree_recurse(PlannerInfo *root, Node *jtnode,
 														 &jtlink);
 				break;
 			case JOIN_LEFT:
+#ifdef NOT_USED					/* see XXX comment above */
 				j->quals = pull_up_sublinks_qual_recurse(root, j->quals,
 														 rightrelids,
 														 &j->rarg);
+#endif
 				break;
 			case JOIN_FULL:
 				/* can't do anything with full-join quals */
 				break;
 			case JOIN_RIGHT:
+#ifdef NOT_USED					/* see XXX comment above */
 				j->quals = pull_up_sublinks_qual_recurse(root, j->quals,
 														 leftrelids,
 														 &j->larg);
+#endif
 				break;
 			default:
 				elog(ERROR, "unrecognized join type: %d",