Fix convert_IN_to_join to properly handle the case where the subselect's

output is not of the same type that's needed for the IN comparison (ie,
where the parser inserted an implicit coercion above the subselect result).
We should record the coerced expression, not just a raw Var referencing
the subselect output, as the quantity that needs to be unique-ified if
we choose to implement the IN as Unique followed by a plain join.

As of 8.3 this error was causing crashes, as seen in bug #4113 from Javier
Hernandez, because the executor was being told to hash or sort the raw
subselect output column using operators appropriate to the coerced type.

In prior versions there was no crash because the executor chose the
hash or sort operators for itself based on the column type it saw.
However, that's still not really right, because what's unique for one data
type might not be unique for another.  In corner cases we could get multiple
outputs of a row that should appear only once, as demonstrated by the
regression test case included in this commit.

However, this patch doesn't apply cleanly to 8.2 or before, and the code
involved has shifted enough over time that I'm hesitant to try to back-patch.
Given the lack of complaints from the field about such corner cases, I think
the bug may not be important enough to risk breaking other things with a
back-patch.

src/backend/optimizer/plan/subselect.c

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/subselect.c,v 1.129 2008/01/17 20:35:27 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/subselect.c,v 1.130 2008/04/21 20:54:15 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -725,10 +725,13 @@ convert_IN_to_join(PlannerInfo *root, SubLink *sublink)
 	Query	   *parse = root->parse;
 	Query	   *subselect = (Query *) sublink->subselect;
 	List	   *in_operators;
+	List	   *left_exprs;
+	List	   *right_exprs;
 	Relids		left_varnos;
 	int			rtindex;
 	RangeTblEntry *rte;
 	RangeTblRef *rtr;
+	List	   *subquery_vars;
 	InClauseInfo *ininfo;
 	Node	   *result;
 
@@ -744,28 +747,37 @@ convert_IN_to_join(PlannerInfo *root, SubLink *sublink)
 		return NULL;
 	if (sublink->testexpr && IsA(sublink->testexpr, OpExpr))
 	{
-		Oid			opno = ((OpExpr *) sublink->testexpr)->opno;
+		OpExpr	   *op = (OpExpr *) sublink->testexpr;
+		Oid			opno = op->opno;
 		List	   *opfamilies;
 		List	   *opstrats;
 
+		if (list_length(op->args) != 2)
+			return NULL;				/* not binary operator? */
 		get_op_btree_interpretation(opno, &opfamilies, &opstrats);
 		if (!list_member_int(opstrats, ROWCOMPARE_EQ))
 			return NULL;
 		in_operators = list_make1_oid(opno);
+		left_exprs = list_make1(linitial(op->args));
+		right_exprs = list_make1(lsecond(op->args));
 	}
 	else if (and_clause(sublink->testexpr))
 	{
 		ListCell   *lc;
 
-		/* OK, but we need to extract the per-column operator OIDs */
-		in_operators = NIL;
+		/* OK, but we need to extract the per-column info */
+		in_operators = left_exprs = right_exprs = NIL;
 		foreach(lc, ((BoolExpr *) sublink->testexpr)->args)
 		{
 			OpExpr	   *op = (OpExpr *) lfirst(lc);
 
 			if (!IsA(op, OpExpr))		/* probably shouldn't happen */
 				return NULL;
+			if (list_length(op->args) != 2)
+				return NULL;			/* not binary operator? */
 			in_operators = lappend_oid(in_operators, op->opno);
+			left_exprs = lappend(left_exprs, linitial(op->args));
+			right_exprs = lappend(right_exprs, lsecond(op->args));
 		}
 	}
 	else
@@ -782,10 +794,13 @@ convert_IN_to_join(PlannerInfo *root, SubLink *sublink)
 	 * The left-hand expressions must contain some Vars of the current query,
 	 * else it's not gonna be a join.
 	 */
-	left_varnos = pull_varnos(sublink->testexpr);
+	left_varnos = pull_varnos((Node *) left_exprs);
 	if (bms_is_empty(left_varnos))
 		return NULL;
 
+	/* ... and the right-hand expressions better not contain Vars at all */
+	Assert(!contain_var_clause((Node *) right_exprs));
+
 	/*
 	 * The combining operators and left-hand expressions mustn't be volatile.
 	 */
@@ -810,6 +825,20 @@ convert_IN_to_join(PlannerInfo *root, SubLink *sublink)
 	rtr->rtindex = rtindex;
 	parse->jointree->fromlist = lappend(parse->jointree->fromlist, rtr);
 
+	/*
+	 * Build a list of Vars representing the subselect outputs.
+	 */
+	subquery_vars = generate_subquery_vars(root,
+										   subselect->targetList,
+										   rtindex);
+
+	/*
+	 * Build the result qual expression, replacing Params with these Vars.
+	 */
+	result = convert_testexpr(root,
+							  sublink->testexpr,
+							  subquery_vars);
+
 	/*
 	 * Now build the InClauseInfo node.
 	 */
@@ -819,24 +848,20 @@ convert_IN_to_join(PlannerInfo *root, SubLink *sublink)
 	ininfo->in_operators = in_operators;
 
 	/*
-	 * ininfo->sub_targetlist is filled with a list of Vars representing the
-	 * subselect outputs.
+	 * ininfo->sub_targetlist must be filled with a list of expressions that
+	 * would need to be unique-ified if we try to implement the IN using a
+	 * regular join to unique-ified subquery output.  This is most easily done
+	 * by applying convert_testexpr to just the RHS inputs of the testexpr
+	 * operators.  That handles cases like type coercions of the subquery
+	 * outputs, clauses dropped due to const-simplification, etc.
 	 */
-	ininfo->sub_targetlist = generate_subquery_vars(root,
-													subselect->targetList,
-													rtindex);
-	Assert(list_length(in_operators) == list_length(ininfo->sub_targetlist));
+	ininfo->sub_targetlist = (List *) convert_testexpr(root,
+													   (Node *) right_exprs,
+													   subquery_vars);
 
 	/* Add the completed node to the query's list */
 	root->in_info_list = lappend(root->in_info_list, ininfo);
 
-	/*
-	 * Build the result qual expression.
-	 */
-	result = convert_testexpr(root,
-							  sublink->testexpr,
-							  ininfo->sub_targetlist);
-
 	return result;
 }
 

src/backend/optimizer/util/pathnode.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/pathnode.c,v 1.142 2008/01/01 19:45:50 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/pathnode.c,v 1.143 2008/04/21 20:54:15 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -898,7 +898,7 @@ create_unique_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath)
 /*
  * translate_sub_tlist - get subquery column numbers represented by tlist
  *
- * The given targetlist should contain only Vars referencing the given relid.
+ * The given targetlist usually contains only Vars referencing the given relid.
  * Extract their varattnos (ie, the column numbers of the subquery) and return
  * as an integer List.
  *