Upgrade cost estimation for joins, per discussion with Bradley Baetz.

Try to model the effect of rescanning input tuples in mergejoins;
account for JOIN_IN short-circuiting where appropriate.  Also, recognize
that mergejoin and hashjoin clauses may now be more than single operator
calls, so we have to charge appropriate execution costs.

src/backend/executor/nodeHashjoin.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeHashjoin.c,v 1.47 2003/01/20 18:54:45 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeHashjoin.c,v 1.48 2003/01/27 20:51:48 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -253,6 +253,13 @@ ExecHashJoin(HashJoinState *node)
 						return result;
 					}
 				}
+
+				/* If we didn't return a tuple, may need to set NeedNewOuter */
+				if (node->js.jointype == JOIN_IN)
+				{
+					node->hj_NeedNewOuter = true;
+					break;		/* out of loop over hash bucket */
+				}
 			}
 		}
 

src/backend/executor/nodeNestloop.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeNestloop.c,v 1.30 2003/01/20 18:54:46 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/executor/nodeNestloop.c,v 1.31 2003/01/27 20:51:48 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -254,6 +254,10 @@ ExecNestLoop(NestLoopState *node)
 					return result;
 				}
 			}
+
+			/* If we didn't return a tuple, may need to set NeedNewOuter */
+			if (node->js.jointype == JOIN_IN)
+				node->nl_NeedNewOuter = true;
 		}
 
 		/*

src/backend/nodes/list.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/list.c,v 1.45 2003/01/24 03:58:34 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/list.c,v 1.46 2003/01/27 20:51:49 tgl Exp $
  *
  * NOTES
  *	  XXX a few of the following functions are duplicated to handle
@@ -639,6 +639,28 @@ set_differencei(List *l1, List *l2)
 	return result;
 }
 
+/*
+ *	set_ptrDifference
+ *
+ *	Same as set_difference, when pointer-equality comparison is sufficient
+ */
+List *
+set_ptrDifference(List *l1, List *l2)
+{
+	List	   *result = NIL;
+	List	   *i;
+
+	if (l2 == NIL)
+		return listCopy(l1);	/* slightly faster path for empty l2 */
+
+	foreach(i, l1)
+	{
+		if (!ptrMember(lfirst(i), l2))
+			result = lappend(result, lfirst(i));
+	}
+	return result;
+}
+
 /*
  * Reverse a list, non-destructively
  */

src/backend/optimizer/path/joinpath.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/optimizer/path/joinpath.c,v 1.76 2003/01/20 18:54:50 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/optimizer/path/joinpath.c,v 1.77 2003/01/27 20:51:51 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -774,13 +774,6 @@ hash_inner_and_outer(Query *root,
  * We examine each restrictinfo clause known for the join to see
  * if it is mergejoinable and involves vars from the two sub-relations
  * currently of interest.
- *
- * Since we currently allow only plain Vars as the left and right sides
- * of mergejoin clauses, this test is relatively simple.  This routine
- * would need to be upgraded to support more-complex expressions
- * as sides of mergejoins.	In theory, we could allow arbitrarily complex
- * expressions in mergejoins, so long as one side uses only vars from one
- * sub-relation and the other side uses only vars from the other.
  */
 static List *
 select_mergejoin_clauses(RelOptInfo *joinrel,
@@ -835,7 +828,9 @@ select_mergejoin_clauses(RelOptInfo *joinrel,
 			continue;			/* not mergejoinable */
 
 		/*
-		 * Check if clause is usable with these input rels.
+		 * Check if clause is usable with these input rels.  All the vars
+		 * needed on each side of the clause must be available from one or
+		 * the other of the input rels.
 		 */
 		if (is_subseti(restrictinfo->left_relids, outerrel->relids) &&
 			is_subseti(restrictinfo->right_relids, innerrel->relids))

src/backend/optimizer/util/pathnode.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/optimizer/util/pathnode.c,v 1.85 2003/01/22 00:07:00 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/optimizer/util/pathnode.c,v 1.86 2003/01/27 20:51:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -743,8 +743,7 @@ create_nestloop_path(Query *root,
 	pathnode->joinrestrictinfo = restrict_clauses;
 	pathnode->path.pathkeys = pathkeys;
 
-	cost_nestloop(&pathnode->path, root, outer_path, inner_path,
-				  restrict_clauses);
+	cost_nestloop(pathnode, root);
 
 	return pathnode;
 }
@@ -816,14 +815,7 @@ create_mergejoin_path(Query *root,
 	pathnode->outersortkeys = outersortkeys;
 	pathnode->innersortkeys = innersortkeys;
 
-	cost_mergejoin(&pathnode->jpath.path,
-				   root,
-				   outer_path,
-				   inner_path,
-				   restrict_clauses,
-				   mergeclauses,
-				   outersortkeys,
-				   innersortkeys);
+	cost_mergejoin(pathnode, root);
 
 	return pathnode;
 }
@@ -861,12 +853,7 @@ create_hashjoin_path(Query *root,
 	pathnode->jpath.path.pathkeys = NIL;
 	pathnode->path_hashclauses = hashclauses;
 
-	cost_hashjoin(&pathnode->jpath.path,
-				  root,
-				  outer_path,
-				  inner_path,
-				  restrict_clauses,
-				  hashclauses);
+	cost_hashjoin(pathnode, root);
 
 	return pathnode;
 }

src/backend/utils/adt/selfuncs.c

@@ -15,7 +15,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/utils/adt/selfuncs.c,v 1.129 2003/01/24 03:58:43 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/utils/adt/selfuncs.c,v 1.130 2003/01/27 20:51:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -2043,6 +2043,8 @@ estimate_num_groups(Query *root, List *groupExprs, double input_rows)
 		varinfos = newvarinfos;
 	} while (varinfos != NIL);
 
+	numdistinct = ceil(numdistinct);
+
 	/* Guard against out-of-range answers */
 	if (numdistinct > input_rows)
 		numdistinct = input_rows;

src/include/nodes/pg_list.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $Id: pg_list.h,v 1.32 2003/01/24 03:58:43 tgl Exp $
+ * $Id: pg_list.h,v 1.33 2003/01/27 20:51:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -138,6 +138,7 @@ extern void set_nth(List *l, int n, void *elem);
 
 extern List *set_difference(List *list1, List *list2);
 extern List *set_differencei(List *list1, List *list2);
+extern List *set_ptrDifference(List *list1, List *list2);
 extern List *lreverse(List *l);
 extern List *set_union(List *list1, List *list2);
 extern List *set_unioni(List *list1, List *list2);

src/include/optimizer/cost.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $Id: cost.h,v 1.50 2003/01/12 22:35:29 tgl Exp $
+ * $Id: cost.h,v 1.51 2003/01/27 20:51:54 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -71,18 +71,9 @@ extern void cost_group(Path *path, Query *root,
 					   int numGroupCols, double numGroups,
 					   Cost input_startup_cost, Cost input_total_cost,
 					   double input_tuples);
-extern void cost_nestloop(Path *path, Query *root,
-			  Path *outer_path, Path *inner_path,
-			  List *restrictlist);
-extern void cost_mergejoin(Path *path, Query *root,
-			   Path *outer_path, Path *inner_path,
-			   List *restrictlist,
-			   List *mergeclauses,
-			   List *outersortkeys, List *innersortkeys);
-extern void cost_hashjoin(Path *path, Query *root,
-			  Path *outer_path, Path *inner_path,
-			  List *restrictlist,
-			  List *hashclauses);
+extern void cost_nestloop(NestPath *path, Query *root);
+extern void cost_mergejoin(MergePath *path, Query *root);
+extern void cost_hashjoin(HashPath *path, Query *root);
 extern void cost_qual_eval(QualCost *cost, List *quals);
 extern void set_baserel_size_estimates(Query *root, RelOptInfo *rel);
 extern void set_joinrel_size_estimates(Query *root, RelOptInfo *rel,

src/test/regress/expected/subselect.out

@@ -58,10 +58,10 @@ SELECT '' AS six, f1 AS "Uncorrelated Field" FROM SUBSELECT_TBL
  six | Uncorrelated Field 
 -----+--------------------
      |                  1
-     |                  1
-     |                  2
      |                  2
      |                  3
+     |                  1
+     |                  2
      |                  3
 (6 rows)
 
@@ -71,10 +71,10 @@ SELECT '' AS six, f1 AS "Uncorrelated Field" FROM SUBSELECT_TBL
  six | Uncorrelated Field 
 -----+--------------------
      |                  1
-     |                  1
-     |                  2
      |                  2
      |                  3
+     |                  1
+     |                  2
      |                  3
 (6 rows)