Teach the system how to use hashing for UNION. (INTERSECT/EXCEPT will follow,

but seem like a separate patch since most of the remaining work is on the
executor side.)  I took the opportunity to push selection of the grouping
operators for set operations into the parser where it belongs.  Otherwise this
is just a small exercise in making prepunion.c consider both alternatives.

As with the recent DISTINCT patch, this means we can UNION on datatypes that
can hash but not sort, and it means that UNION without ORDER BY is no longer
certain to produce sorted output.

src/backend/catalog/dependency.c

@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/catalog/dependency.c,v 1.77 2008/08/02 21:31:59 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/catalog/dependency.c,v 1.78 2008/08/07 01:11:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1597,6 +1597,15 @@ find_expr_references_walker(Node *node,
 		context->rtables = list_delete_first(context->rtables);
 		return result;
 	}
+	else if (IsA(node, SetOperationStmt))
+	{
+		SetOperationStmt *setop = (SetOperationStmt *) node;
+
+		/* we need to look at the groupClauses for operator references */
+		find_expr_references_walker((Node *) setop->groupClauses, context);
+		/* fall through to examine child nodes */
+	}
+
 	return expression_tree_walker(node, find_expr_references_walker,
 								  (void *) context);
 }

src/backend/nodes/copyfuncs.c

@@ -15,7 +15,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/copyfuncs.c,v 1.396 2008/08/02 21:31:59 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/copyfuncs.c,v 1.397 2008/08/07 01:11:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1943,6 +1943,7 @@ _copySetOperationStmt(SetOperationStmt *from)
 	COPY_NODE_FIELD(rarg);
 	COPY_NODE_FIELD(colTypes);
 	COPY_NODE_FIELD(colTypmods);
+	COPY_NODE_FIELD(groupClauses);
 
 	return newnode;
 }

src/backend/nodes/equalfuncs.c

@@ -18,7 +18,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/equalfuncs.c,v 1.325 2008/08/02 21:31:59 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/equalfuncs.c,v 1.326 2008/08/07 01:11:47 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -839,6 +839,7 @@ _equalSetOperationStmt(SetOperationStmt *a, SetOperationStmt *b)
 	COMPARE_NODE_FIELD(rarg);
 	COMPARE_NODE_FIELD(colTypes);
 	COMPARE_NODE_FIELD(colTypmods);
+	COMPARE_NODE_FIELD(groupClauses);
 
 	return true;
 }

src/backend/nodes/outfuncs.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.330 2008/08/05 02:43:17 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.331 2008/08/07 01:11:48 tgl Exp $
  *
  * NOTES
  *	  Every node type that can appear in stored rules' parsetrees *must*
@@ -1780,6 +1780,7 @@ _outSetOperationStmt(StringInfo str, SetOperationStmt *node)
 	WRITE_NODE_FIELD(rarg);
 	WRITE_NODE_FIELD(colTypes);
 	WRITE_NODE_FIELD(colTypmods);
+	WRITE_NODE_FIELD(groupClauses);
 }
 
 static void

src/backend/nodes/readfuncs.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/readfuncs.c,v 1.211 2008/08/02 21:31:59 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/readfuncs.c,v 1.212 2008/08/07 01:11:49 tgl Exp $
  *
  * NOTES
  *	  Path and Plan nodes do not have any readfuncs support, because we
@@ -232,6 +232,7 @@ _readSetOperationStmt(void)
 	READ_NODE_FIELD(rarg);
 	READ_NODE_FIELD(colTypes);
 	READ_NODE_FIELD(colTypmods);
+	READ_NODE_FIELD(groupClauses);
 
 	READ_DONE();
 }

src/backend/optimizer/plan/planner.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planner.c,v 1.239 2008/08/05 16:03:10 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planner.c,v 1.240 2008/08/07 01:11:50 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -68,10 +68,6 @@ static double preprocess_limit(PlannerInfo *root,
 				 double tuple_fraction,
 				 int64 *offset_est, int64 *count_est);
 static void preprocess_groupclause(PlannerInfo *root);
-static Oid *extract_grouping_ops(List *groupClause);
-static AttrNumber *extract_grouping_cols(List *groupClause, List *tlist);
-static bool grouping_is_sortable(List *groupClause);
-static bool grouping_is_hashable(List *groupClause);
 static bool choose_hashed_grouping(PlannerInfo *root,
 					   double tuple_fraction, double limit_tuples,
 					   Path *cheapest_path, Path *sorted_path,
@@ -784,10 +780,9 @@ grouping_planner(PlannerInfo *root, double tuple_fraction)
 
 		/*
 		 * If there's a top-level ORDER BY, assume we have to fetch all the
-		 * tuples.	This might seem too simplistic given all the hackery below
-		 * to possibly avoid the sort ... but a nonzero tuple_fraction is only
-		 * of use to plan_set_operations() when the setop is UNION ALL, and
-		 * the result of UNION ALL is always unsorted.
+		 * tuples.	This might be too simplistic given all the hackery below
+		 * to possibly avoid the sort; but the odds of accurate estimates
+		 * here are pretty low anyway.
 		 */
 		if (parse->sortClause)
 			tuple_fraction = 0.0;
@@ -818,7 +813,8 @@ grouping_planner(PlannerInfo *root, double tuple_fraction)
 		 */
 		Assert(parse->commandType == CMD_SELECT);
 
-		tlist = postprocess_setop_tlist(result_plan->targetlist, tlist);
+		tlist = postprocess_setop_tlist(copyObject(result_plan->targetlist),
+										tlist);
 
 		/*
 		 * Can't handle FOR UPDATE/SHARE here (parser should have checked
@@ -1714,100 +1710,6 @@ preprocess_groupclause(PlannerInfo *root)
 	parse->groupClause = new_groupclause;
 }
 
-/*
- * extract_grouping_ops - make an array of the equality operator OIDs
- *		for a SortGroupClause list
- */
-static Oid *
-extract_grouping_ops(List *groupClause)
-{
-	int			numCols = list_length(groupClause);
-	int			colno = 0;
-	Oid		   *groupOperators;
-	ListCell   *glitem;
-
-	groupOperators = (Oid *) palloc(sizeof(Oid) * numCols);
-
-	foreach(glitem, groupClause)
-	{
-		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
-
-		groupOperators[colno] = groupcl->eqop;
-		Assert(OidIsValid(groupOperators[colno]));
-		colno++;
-	}
-
-	return groupOperators;
-}
-
-/*
- * extract_grouping_cols - make an array of the grouping column resnos
- *		for a SortGroupClause list
- */
-static AttrNumber *
-extract_grouping_cols(List *groupClause, List *tlist)
-{
-	AttrNumber *grpColIdx;
-	int			numCols = list_length(groupClause);
-	int			colno = 0;
-	ListCell   *glitem;
-
-	grpColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols);
-
-	foreach(glitem, groupClause)
-	{
-		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
-		TargetEntry *tle = get_sortgroupclause_tle(groupcl, tlist);
-
-		grpColIdx[colno++] = tle->resno;
-	}
-
-	return grpColIdx;
-}
-
-/*
- * grouping_is_sortable - is it possible to implement grouping list by sorting?
- *
- * This is easy since the parser will have included a sortop if one exists.
- */
-static bool
-grouping_is_sortable(List *groupClause)
-{
-	ListCell   *glitem;
-
-	foreach(glitem, groupClause)
-	{
-		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
-
-		if (!OidIsValid(groupcl->sortop))
-			return false;
-	}
-	return true;
-}
-
-/*
- * grouping_is_hashable - is it possible to implement grouping list by hashing?
- *
- * We assume hashing is OK if the equality operators are marked oprcanhash.
- * (If there isn't actually a supporting hash function, the executor will
- * complain at runtime; but this is a misdeclaration of the operator, not
- * a system bug.)
- */
-static bool
-grouping_is_hashable(List *groupClause)
-{
-	ListCell   *glitem;
-
-	foreach(glitem, groupClause)
-	{
-		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
-
-		if (!op_hashjoinable(groupcl->eqop))
-			return false;
-	}
-	return true;
-}
-
 /*
  * choose_hashed_grouping - should we use hashed grouping?
  *

src/backend/optimizer/util/clauses.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/clauses.c,v 1.260 2008/08/02 21:32:00 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/clauses.c,v 1.261 2008/08/07 01:11:50 tgl Exp $
  *
  * HISTORY
  *	  AUTHOR			DATE			MAJOR EVENT
@@ -3933,6 +3933,8 @@ expression_tree_walker(Node *node,
 					return true;
 				if (walker(setop->rarg, context))
 					return true;
+
+				/* groupClauses are deemed uninteresting */
 			}
 			break;
 		case T_InClauseInfo:
@@ -4535,6 +4537,7 @@ expression_tree_mutator(Node *node,
 				FLATCOPY(newnode, setop, SetOperationStmt);
 				MUTATE(newnode->larg, setop->larg, Node *);
 				MUTATE(newnode->rarg, setop->rarg, Node *);
+				/* We do not mutate groupClauses by default */
 				return (Node *) newnode;
 			}
 			break;

src/backend/optimizer/util/pathnode.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/pathnode.c,v 1.144 2008/08/02 21:32:00 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/pathnode.c,v 1.145 2008/08/07 01:11:50 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -24,7 +24,6 @@
 #include "optimizer/paths.h"
 #include "optimizer/tlist.h"
 #include "parser/parse_expr.h"
-#include "parser/parse_oper.h"
 #include "parser/parsetree.h"
 #include "utils/selfuncs.h"
 #include "utils/lsyscache.h"
@@ -1003,12 +1002,6 @@ query_is_distinct_for(Query *query, List *colnos, List *opids)
 	/*
 	 * UNION, INTERSECT, EXCEPT guarantee uniqueness of the whole output row,
 	 * except with ALL.
-	 *
-	 * XXX this code knows that prepunion.c will adopt the default sort/group
-	 * operators for each column datatype to determine uniqueness.  It'd
-	 * probably be better if these operators were chosen at parse time and
-	 * stored into the parsetree, instead of leaving bits of the planner to
-	 * decide semantics.
 	 */
 	if (query->setOperations)
 	{
@@ -1019,24 +1012,26 @@ query_is_distinct_for(Query *query, List *colnos, List *opids)
 
 		if (!topop->all)
 		{
+			ListCell   *lg;
+
 			/* We're good if all the nonjunk output columns are in colnos */
+			lg = list_head(topop->groupClauses);
 			foreach(l, query->targetList)
 			{
 				TargetEntry *tle = (TargetEntry *) lfirst(l);
-				Oid		tle_eq_opr;
+				SortGroupClause *sgc;
 
 				if (tle->resjunk)
 					continue;	/* ignore resjunk columns */
 
+				/* non-resjunk columns should have grouping clauses */
+				Assert(lg != NULL);
+				sgc = (SortGroupClause *) lfirst(lg);
+				lg = lnext(lg);
+
 				opid = distinct_col_search(tle->resno, colnos, opids);
-				if (!OidIsValid(opid))
-					break;		/* exit early if no match */
-				/* check for compatible semantics */
-				get_sort_group_operators(exprType((Node *) tle->expr),
-										 false, false, false,
-										 NULL, &tle_eq_opr, NULL);
-				if (!OidIsValid(tle_eq_opr) ||
-					!equality_ops_are_compatible(opid, tle_eq_opr))
+				if (!OidIsValid(opid) ||
+					!equality_ops_are_compatible(opid, sgc->eqop))
 					break;		/* exit early if no match */
 			}
 			if (l == NULL)		/* had matches for all? */

src/backend/optimizer/util/tlist.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/tlist.c,v 1.79 2008/08/02 21:32:00 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/tlist.c,v 1.80 2008/08/07 01:11:50 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -18,6 +18,7 @@
 #include "optimizer/tlist.h"
 #include "optimizer/var.h"
 #include "parser/parse_expr.h"
+#include "utils/lsyscache.h"
 
 
 /*****************************************************************************
@@ -202,6 +203,109 @@ get_sortgrouplist_exprs(List *sgClauses, List *targetList)
 }
 
 
+/*****************************************************************************
+ *		Functions to extract data from a list of SortGroupClauses
+ *
+ * These don't really belong in tlist.c, but they are sort of related to the
+ * functions just above, and they don't seem to deserve their own file.
+ *****************************************************************************/
+
+/*
+ * extract_grouping_ops - make an array of the equality operator OIDs
+ *		for a SortGroupClause list
+ */
+Oid *
+extract_grouping_ops(List *groupClause)
+{
+	int			numCols = list_length(groupClause);
+	int			colno = 0;
+	Oid		   *groupOperators;
+	ListCell   *glitem;
+
+	groupOperators = (Oid *) palloc(sizeof(Oid) * numCols);
+
+	foreach(glitem, groupClause)
+	{
+		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
+
+		groupOperators[colno] = groupcl->eqop;
+		Assert(OidIsValid(groupOperators[colno]));
+		colno++;
+	}
+
+	return groupOperators;
+}
+
+/*
+ * extract_grouping_cols - make an array of the grouping column resnos
+ *		for a SortGroupClause list
+ */
+AttrNumber *
+extract_grouping_cols(List *groupClause, List *tlist)
+{
+	AttrNumber *grpColIdx;
+	int			numCols = list_length(groupClause);
+	int			colno = 0;
+	ListCell   *glitem;
+
+	grpColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols);
+
+	foreach(glitem, groupClause)
+	{
+		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
+		TargetEntry *tle = get_sortgroupclause_tle(groupcl, tlist);
+
+		grpColIdx[colno++] = tle->resno;
+	}
+
+	return grpColIdx;
+}
+
+/*
+ * grouping_is_sortable - is it possible to implement grouping list by sorting?
+ *
+ * This is easy since the parser will have included a sortop if one exists.
+ */
+bool
+grouping_is_sortable(List *groupClause)
+{
+	ListCell   *glitem;
+
+	foreach(glitem, groupClause)
+	{
+		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
+
+		if (!OidIsValid(groupcl->sortop))
+			return false;
+	}
+	return true;
+}
+
+/*
+ * grouping_is_hashable - is it possible to implement grouping list by hashing?
+ *
+ * We assume hashing is OK if the equality operators are marked oprcanhash.
+ * (If there isn't actually a supporting hash function, the executor will
+ * complain at runtime; but this is a misdeclaration of the operator, not
+ * a system bug.)
+ */
+bool
+grouping_is_hashable(List *groupClause)
+{
+	ListCell   *glitem;
+
+	foreach(glitem, groupClause)
+	{
+		SortGroupClause *groupcl = (SortGroupClause *) lfirst(glitem);
+
+		if (!op_hashjoinable(groupcl->eqop))
+			return false;
+	}
+	return true;
+}
+
+
+
 /*
  * Does tlist have same output datatypes as listed in colTypes?
  *

src/backend/parser/analyze.c

@@ -17,7 +17,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- *	$PostgreSQL: pgsql/src/backend/parser/analyze.c,v 1.375 2008/08/02 21:32:00 tgl Exp $
+ *	$PostgreSQL: pgsql/src/backend/parser/analyze.c,v 1.376 2008/08/07 01:11:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -33,6 +33,7 @@
 #include "parser/parse_clause.h"
 #include "parser/parse_coerce.h"
 #include "parser/parse_expr.h"
+#include "parser/parse_oper.h"
 #include "parser/parse_relation.h"
 #include "parser/parse_target.h"
 #include "parser/parsetree.h"
@@ -1326,6 +1327,7 @@ transformSetOperationTree(ParseState *pstate, SelectStmt *stmt)
 
 		op->colTypes = NIL;
 		op->colTypmods = NIL;
+		op->groupClauses = NIL;
 		/* don't have a "foreach4", so chase two of the lists by hand */
 		lcm = list_head(lcoltypmods);
 		rcm = list_head(rcoltypmods);
@@ -1349,6 +1351,31 @@ transformSetOperationTree(ParseState *pstate, SelectStmt *stmt)
 			op->colTypes = lappend_oid(op->colTypes, rescoltype);
 			op->colTypmods = lappend_int(op->colTypmods, rescoltypmod);
 
+			/*
+			 * For all cases except UNION ALL, identify the grouping operators
+			 * (and, if available, sorting operators) that will be used to
+			 * eliminate duplicates.
+			 */
+			if (op->op != SETOP_UNION || !op->all)
+			{
+				SortGroupClause *grpcl = makeNode(SortGroupClause);
+				Oid			sortop;
+				Oid			eqop;
+
+				/* determine the eqop and optional sortop */
+				get_sort_group_operators(rescoltype,
+										 false, true, false,
+										 &sortop, &eqop, NULL);
+
+				/* we don't have a tlist yet, so can't assign sortgrouprefs */
+				grpcl->tleSortGroupRef = 0;
+				grpcl->eqop = eqop;
+				grpcl->sortop = sortop;
+				grpcl->nulls_first = false;		/* OK with or without sortop */
+
+				op->groupClauses = lappend(op->groupClauses, grpcl);
+			}
+
 			lcm = lnext(lcm);
 			rcm = lnext(rcm);
 		}

src/backend/parser/parse_clause.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/parser/parse_clause.c,v 1.174 2008/08/05 02:43:17 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/parser/parse_clause.c,v 1.175 2008/08/07 01:11:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -70,6 +70,9 @@ static List *addTargetToSortList(ParseState *pstate, TargetEntry *tle,
 					List *sortlist, List *targetlist,
 					SortByDir sortby_dir, SortByNulls sortby_nulls,
 					List *sortby_opname, bool resolveUnknown);
+static List *addTargetToGroupList(ParseState *pstate, TargetEntry *tle,
+					 List *grouplist, List *targetlist,
+					 bool resolveUnknown);
 
 
 /*
@@ -1355,7 +1358,7 @@ transformGroupClause(ParseState *pstate, List *grouplist,
 		if (!found)
 			result = addTargetToGroupList(pstate, tle,
 										  result, *targetlist,
-										  false, true);
+										  true);
 	}
 
 	return result;
@@ -1456,7 +1459,7 @@ transformDistinctClause(ParseState *pstate,
 			continue;			/* ignore junk */
 		result = addTargetToGroupList(pstate, tle,
 									  result, *targetlist,
-									  false, true);
+									  true);
 	}
 
 	return result;
@@ -1551,7 +1554,7 @@ transformDistinctOnClause(ParseState *pstate, List *distinctlist,
 					 errmsg("SELECT DISTINCT ON expressions must match initial ORDER BY expressions")));
 		result = addTargetToGroupList(pstate, tle,
 									  result, *targetlist,
-									  false, true);
+									  true);
 	}
 
 	return result;
@@ -1679,10 +1682,6 @@ addTargetToSortList(ParseState *pstate, TargetEntry *tle,
  * the TLE is considered "already in the list" if it appears there with any
  * sorting semantics.
  *
- * If requireSortOp is TRUE, we require a sorting operator to be found too.
- * XXX this argument should eventually be obsolete, but for now there are
- * parts of the system that can't support non-sortable grouping lists.
- *
  * If resolveUnknown is TRUE, convert TLEs of type UNKNOWN to TEXT.  If not,
  * do nothing (which implies the search for an equality operator will fail).
  * pstate should be provided if resolveUnknown is TRUE, but can be NULL
@@ -1690,10 +1689,10 @@ addTargetToSortList(ParseState *pstate, TargetEntry *tle,
  *
  * Returns the updated SortGroupClause list.
  */
-List *
+static List *
 addTargetToGroupList(ParseState *pstate, TargetEntry *tle,
 					 List *grouplist, List *targetlist,
-					 bool requireSortOp, bool resolveUnknown)
+					 bool resolveUnknown)
 {
 	Oid			restype = exprType((Node *) tle->expr);
 	Oid			sortop;
@@ -1716,7 +1715,7 @@ addTargetToGroupList(ParseState *pstate, TargetEntry *tle,
 
 		/* determine the eqop and optional sortop */
 		get_sort_group_operators(restype,
-								 requireSortOp, true, false,
+								 false, true, false,
 								 &sortop, &eqop, NULL);
 
 		grpcl->tleSortGroupRef = assignSortGroupRef(tle, targetlist);

src/include/catalog/catversion.h

@@ -37,7 +37,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/catalog/catversion.h,v 1.475 2008/08/05 12:09:30 mha Exp $
+ * $PostgreSQL: pgsql/src/include/catalog/catversion.h,v 1.476 2008/08/07 01:11:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -53,6 +53,6 @@
  */
 
 /*							yyyymmddN */
-#define CATALOG_VERSION_NO	200808051
+#define CATALOG_VERSION_NO	200808061
 
 #endif

src/include/nodes/parsenodes.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/nodes/parsenodes.h,v 1.370 2008/08/02 21:32:00 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/nodes/parsenodes.h,v 1.371 2008/08/07 01:11:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -797,7 +797,12 @@ typedef struct SelectStmt
  * top-level Query node containing the leaf SELECTs as subqueries in its
  * range table.  Its setOperations field shows the tree of set operations,
  * with leaf SelectStmt nodes replaced by RangeTblRef nodes, and internal
- * nodes replaced by SetOperationStmt nodes.
+ * nodes replaced by SetOperationStmt nodes.  Information about the output
+ * column types is added, too.  (Note that the child nodes do not necessarily
+ * produce these types directly, but we've checked that their output types
+ * can be coerced to the output column type.)  Also, if it's not UNION ALL,
+ * information about the types' sort/group semantics is provided in the form
+ * of a SortGroupClause list (same representation as, eg, DISTINCT).
  * ----------------------
  */
 typedef struct SetOperationStmt
@@ -812,6 +817,8 @@ typedef struct SetOperationStmt
 	/* Fields derived during parse analysis: */
 	List	   *colTypes;		/* OID list of output column type OIDs */
 	List	   *colTypmods;		/* integer list of output column typmods */
+	List	   *groupClauses;	/* a list of SortGroupClause's */
+	/* groupClauses is NIL if UNION ALL, but must be set otherwise */
 } SetOperationStmt;
 
 

src/include/optimizer/tlist.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/optimizer/tlist.h,v 1.50 2008/08/02 21:32:01 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/optimizer/tlist.h,v 1.51 2008/08/07 01:11:52 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -32,6 +32,11 @@ extern Node *get_sortgroupclause_expr(SortGroupClause *sgClause,
 extern List *get_sortgrouplist_exprs(List *sgClauses,
 						List *targetList);
 
+extern Oid *extract_grouping_ops(List *groupClause);
+extern AttrNumber *extract_grouping_cols(List *groupClause, List *tlist);
+extern bool grouping_is_sortable(List *groupClause);
+extern bool grouping_is_hashable(List *groupClause);
+
 extern bool tlist_same_datatypes(List *tlist, List *colTypes, bool junkOK);
 
 #endif   /* TLIST_H */

src/include/parser/parse_clause.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/parser/parse_clause.h,v 1.51 2008/08/02 21:32:01 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/parser/parse_clause.h,v 1.52 2008/08/07 01:11:52 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -35,9 +35,6 @@ extern List *transformDistinctClause(ParseState *pstate,
 extern List *transformDistinctOnClause(ParseState *pstate, List *distinctlist,
 						List **targetlist, List *sortClause);
 
-extern List *addTargetToGroupList(ParseState *pstate, TargetEntry *tle,
-					 List *grouplist, List *targetlist,
-					 bool requireSortOp, bool resolveUnknown);
 extern Index assignSortGroupRef(TargetEntry *tle, List *tlist);
 extern bool targetIsInSortList(TargetEntry *tle, Oid sortop, List *sortList);
 

src/test/regress/expected/copyselect.out

@@ -73,7 +73,7 @@ copy (select * from test1 join test2 using (id)) to stdout;
 --
 -- Test UNION SELECT
 --
-copy (select t from test1 where id = 1 UNION select * from v_test1) to stdout;
+copy (select t from test1 where id = 1 UNION select * from v_test1 ORDER BY 1) to stdout;
 a
 v_a
 v_b
@@ -83,7 +83,7 @@ v_e
 --
 -- Test subselect
 --
-copy (select * from (select t from test1 where id = 1 UNION select * from v_test1) t1) to stdout;
+copy (select * from (select t from test1 where id = 1 UNION select * from v_test1 ORDER BY 1) t1) to stdout;
 a
 v_a
 v_b

src/test/regress/expected/union.out

@@ -102,7 +102,7 @@ SELECT 1.1 AS three UNION SELECT 2 UNION SELECT 3;
      3
 (3 rows)
 
-SELECT 1.1::float8 AS two UNION SELECT 2 UNION SELECT 2.0::float8;
+SELECT 1.1::float8 AS two UNION SELECT 2 UNION SELECT 2.0::float8 ORDER BY 1;
  two 
 -----
  1.1
@@ -129,7 +129,8 @@ SELECT 1.1 AS two UNION (SELECT 2 UNION ALL SELECT 2);
 --
 SELECT f1 AS five FROM FLOAT8_TBL
 UNION
-SELECT f1 FROM FLOAT8_TBL;
+SELECT f1 FROM FLOAT8_TBL
+ORDER BY 1;
          five          
 -----------------------
  -1.2345678901234e+200
@@ -158,7 +159,8 @@ SELECT f1 FROM FLOAT8_TBL;
 
 SELECT f1 AS nine FROM FLOAT8_TBL
 UNION
-SELECT f1 FROM INT4_TBL;
+SELECT f1 FROM INT4_TBL
+ORDER BY 1;
          nine          
 -----------------------
  -1.2345678901234e+200
@@ -205,7 +207,8 @@ SELECT f1 FROM INT4_TBL
 
 SELECT CAST(f1 AS char(4)) AS three FROM VARCHAR_TBL
 UNION
-SELECT f1 FROM CHAR_TBL;
+SELECT f1 FROM CHAR_TBL
+ORDER BY 1;
  three 
 -------
  a   
@@ -215,7 +218,8 @@ SELECT f1 FROM CHAR_TBL;
 
 SELECT f1 AS three FROM VARCHAR_TBL
 UNION
-SELECT CAST(f1 AS varchar) FROM CHAR_TBL;
+SELECT CAST(f1 AS varchar) FROM CHAR_TBL
+ORDER BY 1;
  three 
 -------
  a
@@ -242,7 +246,8 @@ SELECT f1 AS five FROM TEXT_TBL
 UNION
 SELECT f1 FROM VARCHAR_TBL
 UNION
-SELECT TRIM(TRAILING FROM f1) FROM CHAR_TBL;
+SELECT TRIM(TRAILING FROM f1) FROM CHAR_TBL
+ORDER BY 1;
        five        
 -------------------
  a

src/test/regress/sql/copyselect.sql

@@ -53,11 +53,11 @@ copy (select * from test1 join test2 using (id)) to stdout;
 --
 -- Test UNION SELECT
 --
-copy (select t from test1 where id = 1 UNION select * from v_test1) to stdout;
+copy (select t from test1 where id = 1 UNION select * from v_test1 ORDER BY 1) to stdout;
 --
 -- Test subselect
 --
-copy (select * from (select t from test1 where id = 1 UNION select * from v_test1) t1) to stdout;
+copy (select * from (select t from test1 where id = 1 UNION select * from v_test1 ORDER BY 1) t1) to stdout;
 --
 -- Test headers, CSV and quotes
 --

src/test/regress/sql/union.sql

@@ -34,7 +34,7 @@ SELECT 1.0::float8 AS two UNION ALL SELECT 1;
 
 SELECT 1.1 AS three UNION SELECT 2 UNION SELECT 3;
 
-SELECT 1.1::float8 AS two UNION SELECT 2 UNION SELECT 2.0::float8;
+SELECT 1.1::float8 AS two UNION SELECT 2 UNION SELECT 2.0::float8 ORDER BY 1;
 
 SELECT 1.1 AS three UNION SELECT 2 UNION ALL SELECT 2;
 
@@ -46,7 +46,8 @@ SELECT 1.1 AS two UNION (SELECT 2 UNION ALL SELECT 2);
 
 SELECT f1 AS five FROM FLOAT8_TBL
 UNION
-SELECT f1 FROM FLOAT8_TBL;
+SELECT f1 FROM FLOAT8_TBL
+ORDER BY 1;
 
 SELECT f1 AS ten FROM FLOAT8_TBL
 UNION ALL
@@ -54,7 +55,8 @@ SELECT f1 FROM FLOAT8_TBL;
 
 SELECT f1 AS nine FROM FLOAT8_TBL
 UNION
-SELECT f1 FROM INT4_TBL;
+SELECT f1 FROM INT4_TBL
+ORDER BY 1;
 
 SELECT f1 AS ten FROM FLOAT8_TBL
 UNION ALL
@@ -68,11 +70,13 @@ SELECT f1 FROM INT4_TBL
 
 SELECT CAST(f1 AS char(4)) AS three FROM VARCHAR_TBL
 UNION
-SELECT f1 FROM CHAR_TBL;
+SELECT f1 FROM CHAR_TBL
+ORDER BY 1;
 
 SELECT f1 AS three FROM VARCHAR_TBL
 UNION
-SELECT CAST(f1 AS varchar) FROM CHAR_TBL;
+SELECT CAST(f1 AS varchar) FROM CHAR_TBL
+ORDER BY 1;
 
 SELECT f1 AS eight FROM VARCHAR_TBL
 UNION ALL
@@ -82,7 +86,8 @@ SELECT f1 AS five FROM TEXT_TBL
 UNION
 SELECT f1 FROM VARCHAR_TBL
 UNION
-SELECT TRIM(TRAILING FROM f1) FROM CHAR_TBL;
+SELECT TRIM(TRAILING FROM f1) FROM CHAR_TBL
+ORDER BY 1;
 
 --
 -- INTERSECT and EXCEPT