Repair bug reported by Wickstrom: backend would crash if WHERE clause

contained a sub-SELECT nested within an AND/OR tree that cnfify()
thought it should rearrange.  Same physical sub-SELECT node could
end up linked into multiple places in resulting expression tree.
This is harmless for most node types, but not for SubLink.
Repair bug by making physical copies of subexpressions that get
logically duplicated by cnfify().  Also, tweak the heuristic that
decides whether it's a good idea to do cnfify() --- we don't really
want that to happen when it would cause multiple copies of a subselect
to be generated, I think.

diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
index 3493bfda24569bd2339821f758cbff49004dc11f..373b05d42fe79c4038871585d38ed28a911fa617 100644 (file)
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *       $Header: /cvsroot/pgsql/src/backend/optimizer/plan/subselect.c,v 1.35 2000/04/12 17:15:22 momjian Exp $
+ *       $Header: /cvsroot/pgsql/src/backend/optimizer/plan/subselect.c,v 1.36 2000/04/14 00:19:16 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -136,6 +136,13 @@ make_subplan(SubLink *slink)
 
        PlannerQueryLevel++;            /* we become child */
 
+       /* Check to see if this node was already processed; if so we have
+        * trouble.  Someday should change tree representation so that we can
+        * cope with multiple links to the same subquery, but for now...
+        */
+       if (subquery == NULL)
+               elog(ERROR, "make_subplan: invalid expression structure (subquery already processed?)");
+
        /*
         * For an EXISTS subplan, tell lower-level planner to expect that only
         * the first tuple will be retrieved.  For ALL and ANY subplans, we
@@ -194,7 +201,7 @@ make_subplan(SubLink *slink)
        node->plan_id = PlannerPlanId++;
        node->rtable = subquery->rtable;
        node->sublink = slink;
-       slink->subselect = NULL;        /* cool ?! */
+       slink->subselect = NULL;        /* cool ?! see error check above! */
 
        /* make parParam list of params coming from current query level */
        foreach(lst, plan->extParam)

diff --git a/src/backend/optimizer/prep/prepqual.c b/src/backend/optimizer/prep/prepqual.c
index fae694dc264dcfd297945e20703938ffd3c550a4..fed6f66f6a17b3662c102720f885f23f969533a7 100644 (file)
--- a/src/backend/optimizer/prep/prepqual.c
+++ b/src/backend/optimizer/prep/prepqual.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *       $Header: /cvsroot/pgsql/src/backend/optimizer/prep/prepqual.c,v 1.24 2000/04/12 17:15:23 momjian Exp $
+ *       $Header: /cvsroot/pgsql/src/backend/optimizer/prep/prepqual.c,v 1.25 2000/04/14 00:19:17 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -656,15 +656,26 @@ or_normalize(List *orlist)
        foreach(temp, distributable->args)
        {
                Expr       *andclause = lfirst(temp);
+               List       *neworlist;
+
+               /*
+                * We are going to insert the orlist into multiple places in the
+                * result expression.  For most expression types, it'd be OK to
+                * just have multiple links to the same subtree, but this fails
+                * badly for SubLinks (and perhaps other cases?).  For safety,
+                * we make a distinct copy for each place the orlist is inserted.
+                */
+               if (lnext(temp) == NIL)
+                       neworlist = orlist;     /* can use original tree at the end */
+               else
+                       neworlist = copyObject(orlist);
 
                /*
                 * pull_ors is needed here in case andclause has a top-level OR.
                 * Then we recursively apply or_normalize, since there might be an
-                * AND subclause in the resulting OR-list. Note: we rely on
-                * pull_ors to build a fresh list, and not damage the given
-                * orlist.
+                * AND subclause in the resulting OR-list.
                 */
-               andclause = or_normalize(pull_ors(lcons(andclause, orlist)));
+               andclause = or_normalize(pull_ors(lcons(andclause, neworlist)));
                andclauses = lappend(andclauses, andclause);
        }
 
@@ -773,15 +784,26 @@ and_normalize(List *andlist)
        foreach(temp, distributable->args)
        {
                Expr       *orclause = lfirst(temp);
+               List       *newandlist;
+
+               /*
+                * We are going to insert the andlist into multiple places in the
+                * result expression.  For most expression types, it'd be OK to
+                * just have multiple links to the same subtree, but this fails
+                * badly for SubLinks (and perhaps other cases?).  For safety,
+                * we make a distinct copy for each place the andlist is inserted.
+                */
+               if (lnext(temp) == NIL)
+                       newandlist = andlist;   /* can use original tree at the end */
+               else
+                       newandlist = copyObject(andlist);
 
                /*
                 * pull_ands is needed here in case orclause has a top-level AND.
                 * Then we recursively apply and_normalize, since there might be
-                * an OR subclause in the resulting AND-list. Note: we rely on
-                * pull_ands to build a fresh list, and not damage the given
-                * andlist.
+                * an OR subclause in the resulting AND-list.
                 */
-               orclause = and_normalize(pull_ands(lcons(orclause, andlist)));
+               orclause = and_normalize(pull_ands(lcons(orclause, newandlist)));
                orclauses = lappend(orclauses, orclause);
        }
 
@@ -932,6 +954,17 @@ count_bool_nodes(Expr *qual,
                count_bool_nodes(get_notclausearg(qual),
                                                 nodes, cnfnodes, dnfnodes);
        }
+       else if (contain_subplans((Node *) qual))
+       {
+               /* charge extra for subexpressions containing sub-SELECTs,
+                * to discourage us from rearranging them in a way that might
+                * generate N copies of a subselect rather than one.  The magic
+                * constant here interacts with the "4x maximum growth" heuristic
+                * in canonicalize_qual().
+                */
+               *nodes = 1.0;
+               *cnfnodes = *dnfnodes = 25.0;
+       }
        else
        {
                /* anything else counts 1 for my purposes */