Refactor planner's pathkeys data structure to create a separate, explicit

representation of equivalence classes of variables.  This is an extensive
rewrite, but it brings a number of benefits:
* planner no longer fails in the presence of "incomplete" operator families
that don't offer operators for every possible combination of datatypes.
* avoid generating and then discarding redundant equality clauses.
* remove bogus assumption that derived equalities always use operators
named "=".
* mergejoins can work with a variety of sort orders (e.g., descending) now,
instead of tying each mergejoinable operator to exactly one sort order.
* better recognition of redundant sort columns.
* can make use of equalities appearing underneath an outer join.

doc/src/sgml/xoper.sgml

-<!-- $PostgreSQL: pgsql/doc/src/sgml/xoper.sgml,v 1.37 2006/12/23 00:43:08 tgl Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/xoper.sgml,v 1.38 2007/01/20 20:45:38 tgl Exp $ -->
 
  <sect1 id="xoper">
   <title>User-Defined Operators</title>
@@ -337,7 +337,7 @@ table1.column1 OP table2.column2
      join will never compare them at all, implicitly assuming that the
      result of the join operator must be false.  So it never makes sense
      to specify <literal>HASHES</literal> for operators that do not represent
-     equality.
+     some form of equality.
     </para>
 
     <para>
@@ -347,7 +347,7 @@ table1.column1 OP table2.column2
      exist yet.  But attempts to use the operator in hash joins will fail
      at run time if no such operator family exists.  The system needs the
      operator family to find the data-type-specific hash function for the
-     operator's input data type.  Of course, you must also supply a suitable
+     operator's input data type.  Of course, you must also create a suitable
      hash function before you can create the operator family.
     </para>
 
@@ -382,8 +382,9 @@ table1.column1 OP table2.column2
      false, never null, for any two nonnull inputs.  If this rule is
      not followed, hash-optimization of <literal>IN</> operations may
      generate wrong results.  (Specifically, <literal>IN</> might return
-     false where the correct answer according to the standard would be null; or it might
-     yield an error complaining that it wasn't prepared for a null result.)
+     false where the correct answer according to the standard would be null;
+     or it might yield an error complaining that it wasn't prepared for a
+     null result.)
     </para>
     </note>
 
@@ -407,19 +408,18 @@ table1.column1 OP table2.column2
      that can only succeed for pairs of values that fall at the
      <quote>same place</>
      in the sort order.  In practice this means that the join operator must
-     behave like equality.  But unlike hash join, where the left and right
-     data types had better be the same (or at least bitwise equivalent),
-     it is possible to merge-join two
+     behave like equality.  But it is possible to merge-join two
      distinct data types so long as they are logically compatible.  For
-     example, the <type>smallint</type>-versus-<type>integer</type> equality operator
-     is merge-joinable.
+     example, the <type>smallint</type>-versus-<type>integer</type>
+     equality operator is merge-joinable.
      We only need sorting operators that will bring both data types into a
      logically compatible sequence.
     </para>
 
     <para>
      To be marked <literal>MERGES</literal>, the join operator must appear
-     in a btree index operator family.  This is not enforced when you create
+     as an equality member of a btree index operator family.
+     This is not enforced when you create
      the operator, since of course the referencing operator family couldn't
      exist yet.  But the operator will not actually be used for merge joins
      unless a matching operator family can be found.  The
@@ -428,30 +428,14 @@ table1.column1 OP table2.column2
     </para>
 
     <para>
-     There are additional restrictions on operators that you mark
-     merge-joinable.  These restrictions are not currently checked by
-     <command>CREATE OPERATOR</command>, but errors may occur when
-     the operator is used if any are not true:
-
-     <itemizedlist>
-      <listitem>
-       <para>
-	A merge-joinable equality operator must have a merge-joinable
-        commutator (itself if the two operand data types are the same, or a related
-        equality operator if they are different).
-       </para>
-      </listitem>
-
-      <listitem>
-       <para>
-        If there is a merge-joinable operator relating any two data types
-	A and B, and another merge-joinable operator relating B to any
-	third data type C, then A and C must also have a merge-joinable
-	operator; in other words, having a merge-joinable operator must
-	be transitive.
-       </para>
-      </listitem>
-     </itemizedlist>
+     A merge-joinable operator must have a commutator (itself if the two
+     operand data types are the same, or a related equality operator
+     if they are different) that appears in the same operator family.
+     If this is not the case, planner errors may occur when the operator
+     is used.  Also, it is a good idea (but not strictly required) for
+     a btree operator family that supports multiple datatypes to provide
+     equality operators for every combination of the datatypes; this
+     allows better optimization.
     </para>
 
     <note>

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 1.361 2007/01/10 18:06:02 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/copyfuncs.c,v 1.362 2007/01/20 20:45:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1284,16 +1284,18 @@ _copyFromExpr(FromExpr *from)
  */
 
 /*
- * _copyPathKeyItem
+ * _copyPathKey
  */
-static PathKeyItem *
-_copyPathKeyItem(PathKeyItem *from)
+static PathKey *
+_copyPathKey(PathKey *from)
 {
-	PathKeyItem *newnode = makeNode(PathKeyItem);
+	PathKey *newnode = makeNode(PathKey);
 
-	COPY_NODE_FIELD(key);
-	COPY_SCALAR_FIELD(sortop);
-	COPY_SCALAR_FIELD(nulls_first);
+	/* EquivalenceClasses are never moved, so just shallow-copy the pointer */
+	COPY_SCALAR_FIELD(pk_eclass);
+	COPY_SCALAR_FIELD(pk_opfamily);
+	COPY_SCALAR_FIELD(pk_strategy);
+	COPY_SCALAR_FIELD(pk_nulls_first);
 
 	return newnode;
 }
@@ -1316,21 +1318,15 @@ _copyRestrictInfo(RestrictInfo *from)
 	COPY_BITMAPSET_FIELD(left_relids);
 	COPY_BITMAPSET_FIELD(right_relids);
 	COPY_NODE_FIELD(orclause);
+	/* EquivalenceClasses are never copied, so shallow-copy the pointers */
+	COPY_SCALAR_FIELD(parent_ec);
 	COPY_SCALAR_FIELD(eval_cost);
 	COPY_SCALAR_FIELD(this_selec);
-	COPY_SCALAR_FIELD(mergejoinoperator);
-	COPY_SCALAR_FIELD(left_sortop);
-	COPY_SCALAR_FIELD(right_sortop);
-	COPY_SCALAR_FIELD(mergeopfamily);
-
-	/*
-	 * Do not copy pathkeys, since they'd not be canonical in a copied query
-	 */
-	newnode->left_pathkey = NIL;
-	newnode->right_pathkey = NIL;
-
-	COPY_SCALAR_FIELD(left_mergescansel);
-	COPY_SCALAR_FIELD(right_mergescansel);
+	COPY_NODE_FIELD(mergeopfamilies);
+	/* EquivalenceClasses are never copied, so shallow-copy the pointers */
+	COPY_SCALAR_FIELD(left_ec);
+	COPY_SCALAR_FIELD(right_ec);
+	COPY_SCALAR_FIELD(outer_is_left);
 	COPY_SCALAR_FIELD(hashjoinoperator);
 	COPY_SCALAR_FIELD(left_bucketsize);
 	COPY_SCALAR_FIELD(right_bucketsize);
@@ -3033,8 +3029,8 @@ copyObject(void *from)
 			/*
 			 * RELATION NODES
 			 */
-		case T_PathKeyItem:
-			retval = _copyPathKeyItem(from);
+		case T_PathKey:
+			retval = _copyPathKey(from);
 			break;
 		case T_RestrictInfo:
 			retval = _copyRestrictInfo(from);

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 1.295 2007/01/10 18:06:03 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/equalfuncs.c,v 1.296 2007/01/20 20:45:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -596,11 +596,27 @@ _equalFromExpr(FromExpr *a, FromExpr *b)
  */
 
 static bool
-_equalPathKeyItem(PathKeyItem *a, PathKeyItem *b)
+_equalPathKey(PathKey *a, PathKey *b)
 {
-	COMPARE_NODE_FIELD(key);
-	COMPARE_SCALAR_FIELD(sortop);
-	COMPARE_SCALAR_FIELD(nulls_first);
+	/*
+	 * This is normally used on non-canonicalized PathKeys, so must chase
+	 * up to the topmost merged EquivalenceClass and see if those are the
+	 * same (by pointer equality).
+	 */
+	EquivalenceClass *a_eclass;
+	EquivalenceClass *b_eclass;
+
+	a_eclass = a->pk_eclass;
+	while (a_eclass->ec_merged)
+		a_eclass = a_eclass->ec_merged;
+	b_eclass = b->pk_eclass;
+	while (b_eclass->ec_merged)
+		b_eclass = b_eclass->ec_merged;
+	if (a_eclass != b_eclass)
+		return false;
+	COMPARE_SCALAR_FIELD(pk_opfamily);
+	COMPARE_SCALAR_FIELD(pk_strategy);
+	COMPARE_SCALAR_FIELD(pk_nulls_first);
 
 	return true;
 }
@@ -2016,8 +2032,8 @@ equal(void *a, void *b)
 			/*
 			 * RELATION NODES
 			 */
-		case T_PathKeyItem:
-			retval = _equalPathKeyItem(a, b);
+		case T_PathKey:
+			retval = _equalPathKey(a, b);
 			break;
 		case T_RestrictInfo:
 			retval = _equalRestrictInfo(a, b);

src/backend/nodes/outfuncs.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.293 2007/01/10 18:06:03 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/outfuncs.c,v 1.294 2007/01/20 20:45:38 tgl Exp $
  *
  * NOTES
  *	  Every node type that can appear in stored rules' parsetrees *must*
@@ -1196,29 +1196,11 @@ _outNestPath(StringInfo str, NestPath *node)
 static void
 _outMergePath(StringInfo str, MergePath *node)
 {
-	int			numCols;
-	int			i;
-
 	WRITE_NODE_TYPE("MERGEPATH");
 
 	_outJoinPathInfo(str, (JoinPath *) node);
 
 	WRITE_NODE_FIELD(path_mergeclauses);
-
-	numCols = list_length(node->path_mergeclauses);
-
-	appendStringInfo(str, " :path_mergeFamilies");
-	for (i = 0; i < numCols; i++)
-		appendStringInfo(str, " %u", node->path_mergeFamilies[i]);
-
-	appendStringInfo(str, " :path_mergeStrategies");
-	for (i = 0; i < numCols; i++)
-		appendStringInfo(str, " %d", node->path_mergeStrategies[i]);
-
-	appendStringInfo(str, " :path_mergeNullsFirst");
-	for (i = 0; i < numCols; i++)
-		appendStringInfo(str, " %d", (int) node->path_mergeNullsFirst[i]);
-
 	WRITE_NODE_FIELD(outersortkeys);
 	WRITE_NODE_FIELD(innersortkeys);
 }
@@ -1241,7 +1223,8 @@ _outPlannerInfo(StringInfo str, PlannerInfo *node)
 	/* NB: this isn't a complete set of fields */
 	WRITE_NODE_FIELD(parse);
 	WRITE_NODE_FIELD(join_rel_list);
-	WRITE_NODE_FIELD(equi_key_list);
+	WRITE_NODE_FIELD(eq_classes);
+	WRITE_NODE_FIELD(canon_pathkeys);
 	WRITE_NODE_FIELD(left_join_clauses);
 	WRITE_NODE_FIELD(right_join_clauses);
 	WRITE_NODE_FIELD(full_join_clauses);
@@ -1284,6 +1267,7 @@ _outRelOptInfo(StringInfo str, RelOptInfo *node)
 	WRITE_NODE_FIELD(subplan);
 	WRITE_NODE_FIELD(baserestrictinfo);
 	WRITE_NODE_FIELD(joininfo);
+	WRITE_BOOL_FIELD(has_eclass_joins);
 	WRITE_BITMAPSET_FIELD(index_outer_relids);
 	WRITE_NODE_FIELD(index_inner_paths);
 }
@@ -1306,13 +1290,48 @@ _outIndexOptInfo(StringInfo str, IndexOptInfo *node)
 }
 
 static void
-_outPathKeyItem(StringInfo str, PathKeyItem *node)
+_outEquivalenceClass(StringInfo str, EquivalenceClass *node)
 {
-	WRITE_NODE_TYPE("PATHKEYITEM");
+	/*
+	 * To simplify reading, we just chase up to the topmost merged EC and
+	 * print that, without bothering to show the merge-ees separately.
+	 */
+	while (node->ec_merged)
+		node = node->ec_merged;
 
-	WRITE_NODE_FIELD(key);
-	WRITE_OID_FIELD(sortop);
-	WRITE_BOOL_FIELD(nulls_first);
+	WRITE_NODE_TYPE("EQUIVALENCECLASS");
+
+	WRITE_NODE_FIELD(ec_opfamilies);
+	WRITE_NODE_FIELD(ec_members);
+	WRITE_NODE_FIELD(ec_sources);
+	WRITE_BITMAPSET_FIELD(ec_relids);
+	WRITE_BOOL_FIELD(ec_has_const);
+	WRITE_BOOL_FIELD(ec_has_volatile);
+	WRITE_BOOL_FIELD(ec_below_outer_join);
+	WRITE_BOOL_FIELD(ec_broken);
+}
+
+static void
+_outEquivalenceMember(StringInfo str, EquivalenceMember *node)
+{
+	WRITE_NODE_TYPE("EQUIVALENCEMEMBER");
+
+	WRITE_NODE_FIELD(em_expr);
+	WRITE_BITMAPSET_FIELD(em_relids);
+	WRITE_BOOL_FIELD(em_is_const);
+	WRITE_BOOL_FIELD(em_is_child);
+	WRITE_OID_FIELD(em_datatype);
+}
+
+static void
+_outPathKey(StringInfo str, PathKey *node)
+{
+	WRITE_NODE_TYPE("PATHKEY");
+
+	WRITE_NODE_FIELD(pk_eclass);
+	WRITE_OID_FIELD(pk_opfamily);
+	WRITE_INT_FIELD(pk_strategy);
+	WRITE_BOOL_FIELD(pk_nulls_first);
 }
 
 static void
@@ -1331,12 +1350,11 @@ _outRestrictInfo(StringInfo str, RestrictInfo *node)
 	WRITE_BITMAPSET_FIELD(left_relids);
 	WRITE_BITMAPSET_FIELD(right_relids);
 	WRITE_NODE_FIELD(orclause);
-	WRITE_OID_FIELD(mergejoinoperator);
-	WRITE_OID_FIELD(left_sortop);
-	WRITE_OID_FIELD(right_sortop);
-	WRITE_OID_FIELD(mergeopfamily);
-	WRITE_NODE_FIELD(left_pathkey);
-	WRITE_NODE_FIELD(right_pathkey);
+	WRITE_NODE_FIELD(parent_ec);
+	WRITE_NODE_FIELD(mergeopfamilies);
+	WRITE_NODE_FIELD(left_ec);
+	WRITE_NODE_FIELD(right_ec);
+	WRITE_BOOL_FIELD(outer_is_left);
 	WRITE_OID_FIELD(hashjoinoperator);
 }
 
@@ -2163,8 +2181,14 @@ _outNode(StringInfo str, void *obj)
 			case T_IndexOptInfo:
 				_outIndexOptInfo(str, obj);
 				break;
-			case T_PathKeyItem:
-				_outPathKeyItem(str, obj);
+			case T_EquivalenceClass:
+				_outEquivalenceClass(str, obj);
+				break;
+			case T_EquivalenceMember:
+				_outEquivalenceMember(str, obj);
+				break;
+			case T_PathKey:
+				_outPathKey(str, obj);
 				break;
 			case T_RestrictInfo:
 				_outRestrictInfo(str, obj);

src/backend/nodes/print.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/nodes/print.c,v 1.82 2007/01/05 22:19:30 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/nodes/print.c,v 1.83 2007/01/20 20:45:38 tgl Exp $
  *
  * HISTORY
  *	  AUTHOR			DATE			MAJOR EVENT
@@ -404,7 +404,7 @@ print_expr(Node *expr, List *rtable)
 
 /*
  * print_pathkeys -
- *	  pathkeys list of list of PathKeyItems
+ *	  pathkeys list of PathKeys
  */
 void
 print_pathkeys(List *pathkeys, List *rtable)
@@ -414,17 +414,26 @@ print_pathkeys(List *pathkeys, List *rtable)
 	printf("(");
 	foreach(i, pathkeys)
 	{
-		List	   *pathkey = (List *) lfirst(i);
+		PathKey	   *pathkey = (PathKey *) lfirst(i);
+		EquivalenceClass *eclass;
 		ListCell   *k;
+		bool		first = true;
+
+		eclass = pathkey->pk_eclass;
+		/* chase up, in case pathkey is non-canonical */
+		while (eclass->ec_merged)
+			eclass = eclass->ec_merged;
 
 		printf("(");
-		foreach(k, pathkey)
+		foreach(k, eclass->ec_members)
 		{
-			PathKeyItem *item = (PathKeyItem *) lfirst(k);
+			EquivalenceMember *mem = (EquivalenceMember *) lfirst(k);
 
-			print_expr(item->key, rtable);
-			if (lnext(k))
+			if (first)
+				first = false;
+			else
 				printf(", ");
+			print_expr((Node *) mem->em_expr, rtable);
 		}
 		printf(")");
 		if (lnext(i))

src/backend/optimizer/path/Makefile

@@ -4,7 +4,7 @@
 #    Makefile for optimizer/path
 #
 # IDENTIFICATION
-#    $PostgreSQL: pgsql/src/backend/optimizer/path/Makefile,v 1.17 2007/01/20 17:16:11 petere Exp $
+#    $PostgreSQL: pgsql/src/backend/optimizer/path/Makefile,v 1.18 2007/01/20 20:45:38 tgl Exp $
 #
 #-------------------------------------------------------------------------
 
@@ -12,7 +12,7 @@ subdir = src/backend/optimizer/path
 top_builddir = ../../../..
 include $(top_builddir)/src/Makefile.global
 
-OBJS = allpaths.o clausesel.o costsize.o indxpath.o \
+OBJS = allpaths.o clausesel.o costsize.o equivclass.o indxpath.o \
        joinpath.o joinrels.o orindxpath.o pathkeys.o tidpath.o
 
 all: SUBSYS.o

src/backend/optimizer/path/allpaths.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.156 2007/01/09 02:14:12 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.157 2007/01/20 20:45:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -325,6 +325,16 @@ set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
 			adjust_appendrel_attrs((Node *) rel->joininfo,
 								   appinfo);
 
+		/*
+		 * We have to make child entries in the EquivalenceClass data
+		 * structures as well.
+		 */
+		if (rel->has_eclass_joins)
+		{
+			add_child_rel_equivalences(root, appinfo, rel, childrel);
+			childrel->has_eclass_joins = true;
+		}
+
 		/*
 		 * Copy the parent's attr_needed data as well, with appropriate
 		 * adjustment of relids and attribute numbers.

src/backend/optimizer/path/costsize.c

@@ -54,7 +54,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/path/costsize.c,v 1.174 2007/01/10 18:06:03 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/path/costsize.c,v 1.175 2007/01/20 20:45:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1258,8 +1258,6 @@ cost_mergejoin(MergePath *path, PlannerInfo *root)
 	Path	   *outer_path = path->jpath.outerjoinpath;
 	Path	   *inner_path = path->jpath.innerjoinpath;
 	List	   *mergeclauses = path->path_mergeclauses;
-	Oid		   *mergeFamilies = path->path_mergeFamilies;
-	int		   *mergeStrategies = path->path_mergeStrategies;
 	List	   *outersortkeys = path->outersortkeys;
 	List	   *innersortkeys = path->innersortkeys;
 	Cost		startup_cost = 0;
@@ -1268,7 +1266,6 @@ cost_mergejoin(MergePath *path, PlannerInfo *root)
 	Selectivity merge_selec;
 	QualCost	merge_qual_cost;
 	QualCost	qp_qual_cost;
-	RestrictInfo *firstclause;
 	double		outer_path_rows = PATH_ROWS(outer_path);
 	double		inner_path_rows = PATH_ROWS(inner_path);
 	double		outer_rows,
@@ -1347,32 +1344,47 @@ cost_mergejoin(MergePath *path, PlannerInfo *root)
 	 * inputs that will actually need to be scanned. We use only the first
 	 * (most significant) merge clause for this purpose.
 	 *
-	 * Since this calculation is somewhat expensive, and will be the same for
-	 * all mergejoin paths associated with the merge clause, we cache the
-	 * results in the RestrictInfo node.  XXX that won't work anymore once
-	 * we support multiple possible orderings!
+	 * XXX mergejoinscansel is a bit expensive, can we cache its results?
 	 */
 	if (mergeclauses && path->jpath.jointype != JOIN_FULL)
 	{
-		firstclause = (RestrictInfo *) linitial(mergeclauses);
-		if (firstclause->left_mergescansel < 0) /* not computed yet? */
+		RestrictInfo *firstclause = (RestrictInfo *) linitial(mergeclauses);
+		List	   *opathkeys;
+		List	   *ipathkeys;
+		PathKey	   *opathkey;
+		PathKey	   *ipathkey;
+		Selectivity leftscansel,
+					rightscansel;
+
+		/* Get the input pathkeys to determine the sort-order details */
+		opathkeys = outersortkeys ? outersortkeys : outer_path->pathkeys;
+		ipathkeys = innersortkeys ? innersortkeys : inner_path->pathkeys;
+		Assert(opathkeys);
+		Assert(ipathkeys);
+		opathkey = (PathKey *) linitial(opathkeys);
+		ipathkey = (PathKey *) linitial(ipathkeys);
+		/* debugging check */
+		if (opathkey->pk_opfamily != ipathkey->pk_opfamily ||
+			opathkey->pk_strategy != ipathkey->pk_strategy ||
+			opathkey->pk_nulls_first != ipathkey->pk_nulls_first)
+			elog(ERROR, "left and right pathkeys do not match in mergejoin");
+
 		mergejoinscansel(root, (Node *) firstclause->clause,
-							 mergeFamilies[0],
-							 mergeStrategies[0],
-							 &firstclause->left_mergescansel,
-							 &firstclause->right_mergescansel);
+						 opathkey->pk_opfamily, opathkey->pk_strategy,
+						 &leftscansel, &rightscansel);
 
-		if (bms_is_subset(firstclause->left_relids, outer_path->parent->relids))
+		if (bms_is_subset(firstclause->left_relids,
+						  outer_path->parent->relids))
 		{
 			/* left side of clause is outer */
-			outerscansel = firstclause->left_mergescansel;
-			innerscansel = firstclause->right_mergescansel;
+			outerscansel = leftscansel;
+			innerscansel = rightscansel;
 		}
 		else
 		{
 			/* left side of clause is inner */
-			outerscansel = firstclause->right_mergescansel;
-			innerscansel = firstclause->left_mergescansel;
+			outerscansel = rightscansel;
+			innerscansel = leftscansel;
 		}
 		if (path->jpath.jointype == JOIN_LEFT)
 			outerscansel = 1.0;

src/backend/optimizer/path/joinrels.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/path/joinrels.c,v 1.83 2007/01/05 22:19:31 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/path/joinrels.c,v 1.84 2007/01/20 20:45:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -72,7 +72,7 @@ make_rels_by_joins(PlannerInfo *root, int level, List **joinrels)
 			other_rels = list_head(joinrels[1]);		/* consider all initial
 														 * rels */
 
-		if (old_rel->joininfo != NIL)
+		if (old_rel->joininfo != NIL || old_rel->has_eclass_joins)
 		{
 			/*
 			 * Note that if all available join clauses for this rel require
@@ -152,7 +152,8 @@ make_rels_by_joins(PlannerInfo *root, int level, List **joinrels)
 			 * outer joins --- then we might have to force a bushy outer
 			 * join.  See have_relevant_joinclause().
 			 */
-			if (old_rel->joininfo == NIL && root->oj_info_list == NIL)
+			if (old_rel->joininfo == NIL && !old_rel->has_eclass_joins &&
+				root->oj_info_list == NIL)
 				continue;
 
 			if (k == other_level)
@@ -251,8 +252,7 @@ make_rels_by_joins(PlannerInfo *root, int level, List **joinrels)
 /*
  * make_rels_by_clause_joins
  *	  Build joins between the given relation 'old_rel' and other relations
- *	  that are mentioned within old_rel's joininfo list (i.e., relations
- *	  that participate in join clauses that 'old_rel' also participates in).
+ *	  that participate in join clauses that 'old_rel' also participates in.
  *	  The join rel nodes are returned in a list.
  *
  * 'old_rel' is the relation entry for the relation to be joined

src/backend/optimizer/plan/planmain.c

@@ -14,7 +14,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planmain.c,v 1.98 2007/01/05 22:19:32 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planmain.c,v 1.99 2007/01/20 20:45:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -110,14 +110,14 @@ query_planner(PlannerInfo *root, List *tlist, double tuple_fraction,
 	 * for "simple" rels.
 	 *
 	 * NOTE: in_info_list and append_rel_list were set up by subquery_planner,
-	 * do not touch here
+	 * do not touch here; eq_classes may contain data already, too.
 	 */
 	root->simple_rel_array_size = list_length(parse->rtable) + 1;
 	root->simple_rel_array = (RelOptInfo **)
 		palloc0(root->simple_rel_array_size * sizeof(RelOptInfo *));
 	root->join_rel_list = NIL;
 	root->join_rel_hash = NULL;
-	root->equi_key_list = NIL;
+	root->canon_pathkeys = NIL;
 	root->left_join_clauses = NIL;
 	root->right_join_clauses = NIL;
 	root->full_join_clauses = NIL;
@@ -165,8 +165,8 @@ query_planner(PlannerInfo *root, List *tlist, double tuple_fraction,
 	 * Examine the targetlist and qualifications, adding entries to baserel
 	 * targetlists for all referenced Vars.  Restrict and join clauses are
 	 * added to appropriate lists belonging to the mentioned relations.  We
-	 * also build lists of equijoined keys for pathkey construction, and form
-	 * a target joinlist for make_one_rel() to work from.
+	 * also build EquivalenceClasses for provably equivalent expressions,
+	 * and form a target joinlist for make_one_rel() to work from.
 	 *
 	 * Note: all subplan nodes will have "flat" (var-only) tlists. This
 	 * implies that all expression evaluations are done at the root of the
@@ -179,16 +179,23 @@ query_planner(PlannerInfo *root, List *tlist, double tuple_fraction,
 	joinlist = deconstruct_jointree(root);
 
 	/*
-	 * Use the completed lists of equijoined keys to deduce any implied but
-	 * unstated equalities (for example, A=B and B=C imply A=C).
+	 * Reconsider any postponed outer-join quals now that we have built up
+	 * equivalence classes.  (This could result in further additions or
+	 * mergings of classes.)
 	 */
-	generate_implied_equalities(root);
+	reconsider_outer_join_clauses(root);
 
 	/*
-	 * We should now have all the pathkey equivalence sets built, so it's now
-	 * possible to convert the requested query_pathkeys to canonical form.
-	 * Also canonicalize the groupClause and sortClause pathkeys for use
-	 * later.
+	 * If we formed any equivalence classes, generate additional restriction
+	 * clauses as appropriate.  (Implied join clauses are formed on-the-fly
+	 * later.)
+	 */
+	generate_base_implied_equalities(root);
+
+	/*
+	 * We have completed merging equivalence sets, so it's now possible to
+	 * convert the requested query_pathkeys to canonical form.  Also
+	 * canonicalize the groupClause and sortClause pathkeys for use later.
 	 */
 	root->query_pathkeys = canonicalize_pathkeys(root, root->query_pathkeys);
 	root->group_pathkeys = canonicalize_pathkeys(root, root->group_pathkeys);

src/backend/optimizer/prep/prepjointree.c

@@ -15,7 +15,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepjointree.c,v 1.45 2007/01/05 22:19:32 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepjointree.c,v 1.46 2007/01/20 20:45:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -292,6 +292,7 @@ pull_up_simple_subquery(PlannerInfo *root, Node *jtnode, RangeTblEntry *rte,
 	 */
 	subroot = makeNode(PlannerInfo);
 	subroot->parse = subquery;
+	subroot->planner_cxt = CurrentMemoryContext;
 	subroot->in_info_list = NIL;
 	subroot->append_rel_list = NIL;
 

src/backend/optimizer/prep/prepunion.c

@@ -22,7 +22,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepunion.c,v 1.135 2007/01/05 22:19:32 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/prep/prepunion.c,v 1.136 2007/01/20 20:45:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1195,10 +1195,8 @@ adjust_appendrel_attrs_mutator(Node *node, AppendRelInfo *context)
 		 */
 		newinfo->eval_cost.startup = -1;
 		newinfo->this_selec = -1;
-		newinfo->left_pathkey = NIL;
-		newinfo->right_pathkey = NIL;
-		newinfo->left_mergescansel = -1;
-		newinfo->right_mergescansel = -1;
+		newinfo->left_ec = NULL;
+		newinfo->right_ec = NULL;
 		newinfo->left_bucketsize = -1;
 		newinfo->right_bucketsize = -1;