Major revision of sort-node handling: push knowledge of query

sort order down into planner, instead of handling it only at the very top
level of the planner.  This fixes many things.  An explicit sort is now
avoided if there is a cheaper alternative (typically an indexscan) not
only for ORDER BY, but also for the internal sort of GROUP BY.  It works
even when there is no other reason (such as a WHERE condition) to consider
the indexscan.  It works for indexes on functions.  It works for indexes
on functions, backwards.  It's just so cool...

CAUTION: I have changed the representation of SortClause nodes, therefore
THIS UPDATE BREAKS STORED RULES.  You will need to initdb.

6 files changed, 81 insertions, 97 deletions

diff --git a/src/backend/nodes/copyfuncs.c b/src/backend/nodes/copyfuncs.c
index 99b1c39c7e6..4895be72379 100644
--- a/src/backend/nodes/copyfuncs.c
+++ b/src/backend/nodes/copyfuncs.c
@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/copyfuncs.c,v 1.91 1999/08/16 02:17:41 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/copyfuncs.c,v 1.92 1999/08/21 03:48:57 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -456,12 +456,6 @@ _copyAgg(Agg *from)
 
 	CopyPlanFields((Plan *) from, (Plan *) newnode);
 
-	/*
-	 * Cannot copy agg list; it must be rebuilt to point to subnodes of
-	 * new node.
-	 */
-	set_agg_tlist_references(newnode);
-
 	return newnode;
 }
 
@@ -474,8 +468,8 @@ _copyGroupClause(GroupClause *from)
 {
 	GroupClause *newnode = makeNode(GroupClause);
 
-	newnode->grpOpoid = from->grpOpoid;
-	newnode->tleGroupref = from->tleGroupref;
+	newnode->tleSortGroupRef = from->tleSortGroupRef;
+	newnode->sortop = from->sortop;
 
 	return newnode;
 }
@@ -567,12 +561,11 @@ _copyResdom(Resdom *from)
 	newnode->resno = from->resno;
 	newnode->restype = from->restype;
 	newnode->restypmod = from->restypmod;
-
 	if (from->resname != NULL)
 		newnode->resname = pstrdup(from->resname);
+	newnode->ressortgroupref = from->ressortgroupref;
 	newnode->reskey = from->reskey;
 	newnode->reskeyop = from->reskeyop;
-	newnode->resgroupref = from->resgroupref;
 	newnode->resjunk = from->resjunk;
 
 	return newnode;
@@ -862,8 +855,8 @@ _copyAggref(Aggref *from)
 	newnode->basetype = from->basetype;
 	newnode->aggtype = from->aggtype;
 	Node_Copy(from, newnode, target);
-	newnode->aggno = from->aggno;
 	newnode->usenulls = from->usenulls;
+	newnode->aggno = from->aggno; /* probably not needed */
 
 	return newnode;
 }
@@ -1345,8 +1338,8 @@ _copySortClause(SortClause *from)
 {
 	SortClause *newnode = makeNode(SortClause);
 
-	Node_Copy(from, newnode, resdom);
-	newnode->opoid = from->opoid;
+	newnode->tleSortGroupRef = from->tleSortGroupRef;
+	newnode->sortop = from->sortop;
 
 	return newnode;
 }
@@ -1398,33 +1391,29 @@ _copyQuery(Query *from)
 	newnode->isBinary = from->isBinary;
 	newnode->isTemp = from->isTemp;
 	newnode->unionall = from->unionall;
-	if (from->uniqueFlag)
-		newnode->uniqueFlag = pstrdup(from->uniqueFlag);
-	Node_Copy(from, newnode, sortClause);
+	newnode->hasAggs = from->hasAggs;
+	newnode->hasSubLinks = from->hasSubLinks;
+
 	Node_Copy(from, newnode, rtable);
 	Node_Copy(from, newnode, targetList);
 	Node_Copy(from, newnode, qual);
+	Node_Copy(from, newnode, rowMark);
 
+	if (from->uniqueFlag)
+		newnode->uniqueFlag = pstrdup(from->uniqueFlag);
+	Node_Copy(from, newnode, sortClause);
 	Node_Copy(from, newnode, groupClause);
 	Node_Copy(from, newnode, havingQual);
 
-	newnode->hasAggs = from->hasAggs;
-	newnode->hasSubLinks = from->hasSubLinks;
-
-	if (from->unionClause)
-	{
-		List	   *ulist,
-				   *temp_list = NIL;
-
-		foreach(ulist, from->unionClause)
-			temp_list = lappend(temp_list, copyObject(lfirst(ulist)));
-		newnode->unionClause = temp_list;
-	}
+	/* why is intersectClause missing? */
+	Node_Copy(from, newnode, unionClause);
 
 	Node_Copy(from, newnode, limitOffset);
 	Node_Copy(from, newnode, limitCount);
 
-	Node_Copy(from, newnode, rowMark);
+	/* we do not copy the planner internal fields: base_rel_list,
+	 * join_rel_list, query_pathkeys.  Not entirely clear if this is right?
+	 */
 
 	return newnode;
 }
diff --git a/src/backend/nodes/equalfuncs.c b/src/backend/nodes/equalfuncs.c
index 1c53fe6aede..d8538a45066 100644
--- a/src/backend/nodes/equalfuncs.c
+++ b/src/backend/nodes/equalfuncs.c
@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/equalfuncs.c,v 1.47 1999/08/16 02:17:41 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/equalfuncs.c,v 1.48 1999/08/21 03:48:57 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -34,14 +34,23 @@ _equalResdom(Resdom *a, Resdom *b)
 		return false;
 	if (a->restypmod != b->restypmod)
 		return false;
-	if (strcmp(a->resname, b->resname) != 0)
+	if (a->resname && b->resname)
+	{
+		if (strcmp(a->resname, b->resname) != 0)
+			return false;
+	}
+	else
+	{
+		/* must both be null to be equal */
+		if (a->resname != b->resname)
+			return false;
+	}
+	if (a->ressortgroupref != b->ressortgroupref)
 		return false;
 	if (a->reskey != b->reskey)
 		return false;
 	if (a->reskeyop != b->reskeyop)
 		return false;
-	if (a->resgroupref != b->resgroupref)
-		return false;
 	/* we ignore resjunk flag ... is this correct? */
 
 	return true;
@@ -208,10 +217,9 @@ _equalAggref(Aggref *a, Aggref *b)
 		return false;
 	if (!equal(a->target, b->target))
 		return false;
-	if (a->aggno != b->aggno)
-		return false;
 	if (a->usenulls != b->usenulls)
 		return false;
+	/* ignore aggno, which is only a private field for the executor */
 	return true;
 }
 
@@ -503,6 +511,14 @@ _equalQuery(Query *a, Query *b)
 		return false;
 	if (a->hasSubLinks != b->hasSubLinks)
 		return false;
+	if (!equal(a->rtable, b->rtable))
+		return false;
+	if (!equal(a->targetList, b->targetList))
+		return false;
+	if (!equal(a->qual, b->qual))
+		return false;
+	if (!equal(a->rowMark, b->rowMark))
+		return false;
 	if (a->uniqueFlag && b->uniqueFlag)
 	{
 		if (strcmp(a->uniqueFlag, b->uniqueFlag) != 0)
@@ -515,14 +531,6 @@ _equalQuery(Query *a, Query *b)
 	}
 	if (!equal(a->sortClause, b->sortClause))
 		return false;
-	if (!equal(a->rtable, b->rtable))
-		return false;
-	if (!equal(a->targetList, b->targetList))
-		return false;
-	if (!equal(a->qual, b->qual))
-		return false;
-	if (!equal(a->rowMark, b->rowMark))
-		return false;
 	if (!equal(a->groupClause, b->groupClause))
 		return false;
 	if (!equal(a->havingQual, b->havingQual))
@@ -537,9 +545,9 @@ _equalQuery(Query *a, Query *b)
 		return false;
 
 	/*
-	 * We do not check the internal-to-the-planner fields base_rel_list
-	 * and join_rel_list.  They might not be set yet, and in any case they
-	 * should be derivable from the other fields.
+	 * We do not check the internal-to-the-planner fields: base_rel_list,
+	 * join_rel_list, query_pathkeys.  They might not be set yet, and
+	 * in any case they should be derivable from the other fields.
 	 */
 	return true;
 }
diff --git a/src/backend/nodes/freefuncs.c b/src/backend/nodes/freefuncs.c
index 11f92b28d05..cad49be76bd 100644
--- a/src/backend/nodes/freefuncs.c
+++ b/src/backend/nodes/freefuncs.c
@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/Attic/freefuncs.c,v 1.25 1999/08/16 02:17:42 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/Attic/freefuncs.c,v 1.26 1999/08/21 03:48:57 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -371,8 +371,6 @@ _freeAgg(Agg *node)
 {
 	FreePlanFields((Plan *) node);
 
-	freeList(node->aggs);
-
 	pfree(node);
 }
 
@@ -964,8 +962,6 @@ _freeRowMark(RowMark *node)
 static void
 _freeSortClause(SortClause *node)
 {
-	freeObject(node->resdom);
-
 	pfree(node);
 }
 
@@ -1000,19 +996,22 @@ _freeQuery(Query *node)
 	}
 	if (node->into)
 		pfree(node->into);
+	freeObject(node->rtable);
+	freeObject(node->targetList);
+	freeObject(node->qual);
+	freeObject(node->rowMark);
 	if (node->uniqueFlag)
 		pfree(node->uniqueFlag);
 
 	freeObject(node->sortClause);
-	freeObject(node->rtable);
-	freeObject(node->targetList);
-	freeObject(node->qual);
 	freeObject(node->groupClause);
 	freeObject(node->havingQual);
+	/* why not intersectClause? */
 	freeObject(node->unionClause);
 	freeObject(node->limitOffset);
 	freeObject(node->limitCount);
-	freeObject(node->rowMark);
+
+	/* XXX should we be freeing the planner internal fields? */
 
 	pfree(node);
 }
diff --git a/src/backend/nodes/makefuncs.c b/src/backend/nodes/makefuncs.c
index d64fdd8c462..26ebed1d458 100644
--- a/src/backend/nodes/makefuncs.c
+++ b/src/backend/nodes/makefuncs.c
@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/makefuncs.c,v 1.16 1999/07/15 22:39:17 momjian Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/makefuncs.c,v 1.17 1999/08/21 03:48:58 tgl Exp $
  *
  * NOTES
  *	  Creator functions in POSTGRES 4.2 are generated automatically. Most of
@@ -102,9 +102,12 @@ makeResdom(AttrNumber resno,
 	resdom->restype = restype;
 	resdom->restypmod = restypmod;
 	resdom->resname = resname;
+	/* For historical reasons, ressortgroupref defaults to 0 while
+	 * reskey/reskeyop are passed in explicitly.  This is pretty silly.
+	 */
+	resdom->ressortgroupref = 0;
 	resdom->reskey = reskey;
 	resdom->reskeyop = reskeyop;
-	resdom->resgroupref = 0;
 	resdom->resjunk = resjunk;
 	return resdom;
 }
diff --git a/src/backend/nodes/outfuncs.c b/src/backend/nodes/outfuncs.c
index 39b0625d933..226dbb1cef0 100644
--- a/src/backend/nodes/outfuncs.c
+++ b/src/backend/nodes/outfuncs.c
@@ -5,7 +5,7 @@
  *
  * Copyright (c) 1994, Regents of the University of California
  *
- *	$Id: outfuncs.c,v 1.93 1999/08/16 02:17:42 tgl Exp $
+ *	$Id: outfuncs.c,v 1.94 1999/08/21 03:48:58 tgl Exp $
  *
  * NOTES
  *	  Every (plan) node in POSTGRES has an associated "out" routine which
@@ -237,18 +237,15 @@ _outQuery(StringInfo str, Query *node)
 static void
 _outSortClause(StringInfo str, SortClause *node)
 {
-	appendStringInfo(str, " SORTCLAUSE :resdom ");
-	_outNode(str, node->resdom);
-
-	appendStringInfo(str, " :opoid %u ", node->opoid);
+	appendStringInfo(str, " SORTCLAUSE :tleSortGroupRef %d :sortop %u ",
+					 node->tleSortGroupRef, node->sortop);
 }
 
 static void
 _outGroupClause(StringInfo str, GroupClause *node)
 {
-	appendStringInfo(str, " GROUPCLAUSE :grpOpoid %u :tleGroupref %d",
-					 node->grpOpoid,
-					 node->tleGroupref);
+	appendStringInfo(str, " GROUPCLAUSE :tleSortGroupRef %d :sortop %u ",
+					 node->tleSortGroupRef, node->sortop);
 }
 
 /*
@@ -482,9 +479,6 @@ _outAgg(StringInfo str, Agg *node)
 
 	appendStringInfo(str, " AGG ");
 	_outPlanInfo(str, (Plan *) node);
-
-	appendStringInfo(str, " :aggs ");
-	_outNode(str, node->aggs);
 }
 
 static void
@@ -549,8 +543,8 @@ _outResdom(StringInfo str, Resdom *node)
 					 node->reskey,
 					 node->reskeyop);
 
-	appendStringInfo(str, " :resgroupref %d :resjunk %s ",
-					 node->resgroupref,
+	appendStringInfo(str, " :ressortgroupref %d :resjunk %s ",
+					 node->ressortgroupref,
 					 node->resjunk ? "true" : "false");
 }
 
@@ -665,8 +659,7 @@ _outAggref(StringInfo str, Aggref *node)
 					 node->aggtype);
 	_outNode(str, node->target);
 
-	appendStringInfo(str, ":aggno %d :usenulls %s",
-					 node->aggno,
+	appendStringInfo(str, " :usenulls %s ",
 					 node->usenulls ? "true" : "false");
 }
 
diff --git a/src/backend/nodes/readfuncs.c b/src/backend/nodes/readfuncs.c
index 98385dd96f6..588528daa1d 100644
--- a/src/backend/nodes/readfuncs.c
+++ b/src/backend/nodes/readfuncs.c
@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/nodes/readfuncs.c,v 1.72 1999/08/16 02:17:43 tgl Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/nodes/readfuncs.c,v 1.73 1999/08/21 03:48:58 tgl Exp $
  *
  * NOTES
  *	  Most of the read functions for plan nodes are tested. (In fact, they
@@ -184,12 +184,13 @@ _readSortClause()
 
 	local_node = makeNode(SortClause);
 
-	token = lsptok(NULL, &length);		/* skip the :resdom */
-	local_node->resdom = nodeRead(true);
+	token = lsptok(NULL, &length);		/* skip :tleSortGroupRef */
+	token = lsptok(NULL, &length);		/* get tleSortGroupRef */
+	local_node->tleSortGroupRef = strtoul(token, NULL, 10);
 
-	token = lsptok(NULL, &length);		/* skip :opoid */
-	token = lsptok(NULL, &length);		/* get opoid */
-	local_node->opoid = strtoul(token, NULL, 10);
+	token = lsptok(NULL, &length);		/* skip :sortop */
+	token = lsptok(NULL, &length);		/* get sortop */
+	local_node->sortop = strtoul(token, NULL, 10);
 
 	return local_node;
 }
@@ -207,13 +208,13 @@ _readGroupClause()
 
 	local_node = makeNode(GroupClause);
 
-	token = lsptok(NULL, &length);		/* skip :grpOpoid */
-	token = lsptok(NULL, &length);		/* get grpOpoid */
-	local_node->grpOpoid = strtoul(token, NULL, 10);
+	token = lsptok(NULL, &length);		/* skip :tleSortGroupRef */
+	token = lsptok(NULL, &length);		/* get tleSortGroupRef */
+	local_node->tleSortGroupRef = strtoul(token, NULL, 10);
 
-	token = lsptok(NULL, &length);		/* skip :tleGroupref */
-	token = lsptok(NULL, &length);		/* get tleGroupref */
-	local_node->tleGroupref = strtoul(token, NULL, 10);
+	token = lsptok(NULL, &length);		/* skip :sortop */
+	token = lsptok(NULL, &length);		/* get sortop */
+	local_node->sortop = strtoul(token, NULL, 10);
 
 	return local_node;
 }
@@ -600,15 +601,10 @@ static Agg *
 _readAgg()
 {
 	Agg		   *local_node;
-	char	   *token;
-	int			length;
 
 	local_node = makeNode(Agg);
 	_getPlan((Plan *) local_node);
 
-	token = lsptok(NULL, &length);		/* eat :agg */
-	local_node->aggs = nodeRead(true);	/* now read it */
-
 	return local_node;
 }
 
@@ -712,9 +708,9 @@ _readResdom()
 	token = lsptok(NULL, &length);		/* get reskeyop */
 	local_node->reskeyop = (Oid) atol(token);
 
-	token = lsptok(NULL, &length);		/* eat :resgroupref */
-	token = lsptok(NULL, &length);		/* get resgroupref */
-	local_node->resgroupref = strtoul(token, NULL, 10);
+	token = lsptok(NULL, &length);		/* eat :ressortgroupref */
+	token = lsptok(NULL, &length);		/* get ressortgroupref */
+	local_node->ressortgroupref = strtoul(token, NULL, 10);
 
 	token = lsptok(NULL, &length);		/* eat :resjunk */
 	token = lsptok(NULL, &length);		/* get resjunk */
@@ -1163,10 +1159,6 @@ _readAggref()
 	token = lsptok(NULL, &length);		/* eat :target */
 	local_node->target = nodeRead(true);		/* now read it */
 
-	token = lsptok(NULL, &length);		/* eat :aggno */
-	token = lsptok(NULL, &length);		/* get aggno */
-	local_node->aggno = atoi(token);
-
 	token = lsptok(NULL, &length);		/* eat :usenulls */
 	token = lsptok(NULL, &length);		/* get usenulls */
 	local_node->usenulls = (token[0] == 't') ? true : false;