Adjust parser so that 'x NOT IN (subselect)' is converted to

'NOT (x IN (subselect))', that is 'NOT (x = ANY (subselect))',
rather than 'x <> ALL (subselect)' as we formerly did.  This
opens the door to optimizing NOT IN the same way as IN, whereas
there's no hope of optimizing the expression using <>.  Also,
convert 'x <> ALL (subselect)' to the NOT(IN) style, so that
the optimization will be available when processing rules dumped
by older Postgres versions.
initdb forced due to small change in SubLink node representation.
This commit is contained in:
Tom Lane 2003-01-09 20:50:53 +00:00
parent 59779c81ba
commit 6bc61fc046
11 changed files with 111 additions and 56 deletions

View File

@ -7,7 +7,7 @@
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/executor/nodeSubplan.c,v 1.40 2002/12/26 22:37:42 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/executor/nodeSubplan.c,v 1.41 2003/01/09 20:50:50 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -37,7 +37,7 @@ ExecSubPlan(SubPlanState *node,
SubPlan *subplan = (SubPlan *) node->xprstate.expr;
PlanState *planstate = node->planstate;
SubLinkType subLinkType = subplan->subLinkType;
bool useor = subplan->useor;
bool useOr = subplan->useOr;
MemoryContext oldcontext;
TupleTableSlot *slot;
Datum result;
@ -84,7 +84,7 @@ ExecSubPlan(SubPlanState *node,
* For all sublink types except EXPR_SUBLINK, the result is boolean as
* are the results of the combining operators. We combine results
* within a tuple (if there are multiple columns) using OR semantics
* if "useor" is true, AND semantics if not. We then combine results
* if "useOr" is true, AND semantics if not. We then combine results
* across tuples (if the subplan produces more than one) using OR
* semantics for ANY_SUBLINK or AND semantics for ALL_SUBLINK.
* (MULTIEXPR_SUBLINK doesn't allow multiple tuples from the subplan.)
@ -107,7 +107,7 @@ ExecSubPlan(SubPlanState *node,
{
HeapTuple tup = slot->val;
TupleDesc tdesc = slot->ttc_tupleDescriptor;
Datum rowresult = BoolGetDatum(!useor);
Datum rowresult = BoolGetDatum(!useOr);
bool rownull = false;
int col = 1;
@ -212,7 +212,7 @@ ExecSubPlan(SubPlanState *node,
rowresult = expresult;
rownull = expnull;
}
else if (useor)
else if (useOr)
{
/* combine within row per OR semantics */
if (expnull)

View File

@ -15,7 +15,7 @@
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/nodes/copyfuncs.c,v 1.233 2002/12/14 00:17:50 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/nodes/copyfuncs.c,v 1.234 2003/01/09 20:50:50 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -809,7 +809,8 @@ _copySubLink(SubLink *from)
SubLink *newnode = makeNode(SubLink);
COPY_SCALAR_FIELD(subLinkType);
COPY_SCALAR_FIELD(useor);
COPY_SCALAR_FIELD(operIsEquals);
COPY_SCALAR_FIELD(useOr);
COPY_NODE_FIELD(lefthand);
COPY_NODE_FIELD(oper);
COPY_NODE_FIELD(subselect);
@ -826,7 +827,7 @@ _copySubPlan(SubPlan *from)
SubPlan *newnode = makeNode(SubPlan);
COPY_SCALAR_FIELD(subLinkType);
COPY_SCALAR_FIELD(useor);
COPY_SCALAR_FIELD(useOr);
COPY_NODE_FIELD(oper);
COPY_NODE_FIELD(plan);
COPY_SCALAR_FIELD(plan_id);

View File

@ -18,7 +18,7 @@
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/nodes/equalfuncs.c,v 1.177 2002/12/14 00:17:51 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/nodes/equalfuncs.c,v 1.178 2003/01/09 20:50:50 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -287,7 +287,8 @@ static bool
_equalSubLink(SubLink *a, SubLink *b)
{
COMPARE_SCALAR_FIELD(subLinkType);
COMPARE_SCALAR_FIELD(useor);
COMPARE_SCALAR_FIELD(operIsEquals);
COMPARE_SCALAR_FIELD(useOr);
COMPARE_NODE_FIELD(lefthand);
COMPARE_NODE_FIELD(oper);
COMPARE_NODE_FIELD(subselect);
@ -299,7 +300,7 @@ static bool
_equalSubPlan(SubPlan *a, SubPlan *b)
{
COMPARE_SCALAR_FIELD(subLinkType);
COMPARE_SCALAR_FIELD(useor);
COMPARE_SCALAR_FIELD(useOr);
COMPARE_NODE_FIELD(oper);
/* should compare plans, but have to settle for comparing plan IDs */
COMPARE_SCALAR_FIELD(plan_id);

View File

@ -8,7 +8,7 @@
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/nodes/outfuncs.c,v 1.190 2002/12/14 00:17:52 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/nodes/outfuncs.c,v 1.191 2003/01/09 20:50:50 tgl Exp $
*
* NOTES
* Every node type that can appear in stored rules' parsetrees *must*
@ -658,7 +658,8 @@ _outSubLink(StringInfo str, SubLink *node)
WRITE_NODE_TYPE("SUBLINK");
WRITE_ENUM_FIELD(subLinkType, SubLinkType);
WRITE_BOOL_FIELD(useor);
WRITE_BOOL_FIELD(operIsEquals);
WRITE_BOOL_FIELD(useOr);
WRITE_NODE_FIELD(lefthand);
WRITE_NODE_FIELD(oper);
WRITE_NODE_FIELD(subselect);
@ -670,7 +671,7 @@ _outSubPlan(StringInfo str, SubPlan *node)
WRITE_NODE_TYPE("SUBPLAN");
WRITE_ENUM_FIELD(subLinkType, SubLinkType);
WRITE_BOOL_FIELD(useor);
WRITE_BOOL_FIELD(useOr);
WRITE_NODE_FIELD(oper);
WRITE_NODE_FIELD(plan);
WRITE_INT_FIELD(plan_id);

View File

@ -8,7 +8,7 @@
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/nodes/readfuncs.c,v 1.144 2002/12/14 00:17:54 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/nodes/readfuncs.c,v 1.145 2003/01/09 20:50:51 tgl Exp $
*
* NOTES
* Path and Plan nodes do not have any readfuncs support, because we
@ -531,7 +531,8 @@ _readSubLink(void)
READ_LOCALS(SubLink);
READ_ENUM_FIELD(subLinkType, SubLinkType);
READ_BOOL_FIELD(useor);
READ_BOOL_FIELD(operIsEquals);
READ_BOOL_FIELD(useOr);
READ_NODE_FIELD(lefthand);
READ_NODE_FIELD(oper);
READ_NODE_FIELD(subselect);

View File

@ -7,7 +7,7 @@
* Portions Copyright (c) 1994, Regents of the University of California
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/optimizer/plan/subselect.c,v 1.61 2002/12/14 00:17:55 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/optimizer/plan/subselect.c,v 1.62 2003/01/09 20:50:51 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -225,7 +225,7 @@ make_subplan(SubLink *slink, List *lefthand)
* Fill in other fields of the SubPlan node.
*/
node->subLinkType = slink->subLinkType;
node->useor = slink->useor;
node->useOr = slink->useOr;
node->oper = NIL;
node->setParam = NIL;
node->parParam = NIL;
@ -283,7 +283,7 @@ make_subplan(SubLink *slink, List *lefthand)
&node->setParam);
PlannerInitPlan = lappend(PlannerInitPlan, node);
if (length(oper) > 1)
result = (Node *) (node->useor ? make_orclause(oper) :
result = (Node *) (node->useOr ? make_orclause(oper) :
make_andclause(oper));
else
result = (Node *) lfirst(oper);

View File

@ -11,7 +11,7 @@
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/parser/gram.y,v 2.391 2003/01/08 00:22:27 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/parser/gram.y,v 2.392 2003/01/09 20:50:51 tgl Exp $
*
* HISTORY
* AUTHOR DATE MAJOR EVENT
@ -5420,28 +5420,30 @@ opt_interval:
/* Expressions using row descriptors
* Define row_descriptor to allow yacc to break the reduce/reduce conflict
* with singleton expressions. Use SQL99's ROW keyword to allow rows of
* one element.
* with singleton expressions. Use SQL99's ROW keyword to allow rows of
* one element.
*/
r_expr: row IN_P select_with_parens
{
SubLink *n = makeNode(SubLink);
n->lefthand = $1;
n->oper = (List *) makeSimpleA_Expr(OP, "=", NULL, NULL);
n->useor = FALSE;
n->subLinkType = ANY_SUBLINK;
/* operIsEquals and useOr will be set later */
n->subselect = $3;
$$ = (Node *)n;
}
| row NOT IN_P select_with_parens
{
/* Make an IN node */
SubLink *n = makeNode(SubLink);
n->lefthand = $1;
n->oper = (List *) makeSimpleA_Expr(OP, "<>", NULL, NULL);
n->useor = TRUE;
n->subLinkType = ALL_SUBLINK;
n->oper = (List *) makeSimpleA_Expr(OP, "=", NULL, NULL);
n->subLinkType = ANY_SUBLINK;
/* operIsEquals and useOr will be set later */
n->subselect = $4;
$$ = (Node *)n;
/* Stick a NOT on top */
$$ = (Node *) makeA_Expr(NOT, NIL, NULL, (Node *) n);
}
| row qual_all_Op sub_type select_with_parens
%prec Op
@ -5449,11 +5451,8 @@ r_expr: row IN_P select_with_parens
SubLink *n = makeNode(SubLink);
n->lefthand = $1;
n->oper = (List *) makeA_Expr(OP, $2, NULL, NULL);
if (strcmp(strVal(llast($2)), "<>") == 0)
n->useor = TRUE;
else
n->useor = FALSE;
n->subLinkType = $3;
/* operIsEquals and useOr will be set later */
n->subselect = $4;
$$ = (Node *)n;
}
@ -5463,11 +5462,8 @@ r_expr: row IN_P select_with_parens
SubLink *n = makeNode(SubLink);
n->lefthand = $1;
n->oper = (List *) makeA_Expr(OP, $2, NULL, NULL);
if (strcmp(strVal(llast($2)), "<>") == 0)
n->useor = TRUE;
else
n->useor = FALSE;
n->subLinkType = MULTIEXPR_SUBLINK;
/* operIsEquals and useOr will be set later */
n->subselect = $3;
$$ = (Node *)n;
}
@ -5850,8 +5846,8 @@ a_expr: c_expr { $$ = $1; }
SubLink *n = (SubLink *)$3;
n->lefthand = makeList1($1);
n->oper = (List *) makeSimpleA_Expr(OP, "=", NULL, NULL);
n->useor = FALSE;
n->subLinkType = ANY_SUBLINK;
/* operIsEquals and useOr will be set later */
$$ = (Node *)n;
}
else
@ -5875,12 +5871,14 @@ a_expr: c_expr { $$ = $1; }
/* in_expr returns a SubLink or a list of a_exprs */
if (IsA($4, SubLink))
{
/* Make an IN node */
SubLink *n = (SubLink *)$4;
n->lefthand = makeList1($1);
n->oper = (List *) makeSimpleA_Expr(OP, "<>", NULL, NULL);
n->useor = FALSE;
n->subLinkType = ALL_SUBLINK;
$$ = (Node *)n;
n->oper = (List *) makeSimpleA_Expr(OP, "=", NULL, NULL);
n->subLinkType = ANY_SUBLINK;
/* operIsEquals and useOr will be set later */
/* Stick a NOT on top */
$$ = (Node *) makeA_Expr(NOT, NIL, NULL, (Node *) n);
}
else
{
@ -5903,8 +5901,8 @@ a_expr: c_expr { $$ = $1; }
SubLink *n = makeNode(SubLink);
n->lefthand = makeList1($1);
n->oper = (List *) makeA_Expr(OP, $2, NULL, NULL);
n->useor = FALSE; /* doesn't matter since only one col */
n->subLinkType = $3;
/* operIsEquals and useOr will be set later */
n->subselect = $4;
$$ = (Node *)n;
}
@ -6447,7 +6445,6 @@ c_expr: columnref { $$ = (Node *) $1; }
SubLink *n = makeNode(SubLink);
n->lefthand = NIL;
n->oper = NIL;
n->useor = FALSE;
n->subLinkType = EXPR_SUBLINK;
n->subselect = $1;
$$ = (Node *)n;
@ -6457,7 +6454,6 @@ c_expr: columnref { $$ = (Node *) $1; }
SubLink *n = makeNode(SubLink);
n->lefthand = NIL;
n->oper = NIL;
n->useor = FALSE;
n->subLinkType = EXISTS_SUBLINK;
n->subselect = $2;
$$ = (Node *)n;

View File

@ -8,7 +8,7 @@
*
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/parser/parse_expr.c,v 1.138 2002/12/27 20:06:19 tgl Exp $
* $Header: /cvsroot/pgsql/src/backend/parser/parse_expr.c,v 1.139 2003/01/09 20:50:52 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -367,6 +367,8 @@ transformExpr(ParseState *pstate, Node *expr)
*/
sublink->lefthand = NIL;
sublink->oper = NIL;
sublink->operIsEquals = FALSE;
sublink->useOr = FALSE;
}
else if (sublink->subLinkType == EXPR_SUBLINK)
{
@ -391,27 +393,60 @@ transformExpr(ParseState *pstate, Node *expr)
*/
sublink->lefthand = NIL;
sublink->oper = NIL;
sublink->operIsEquals = FALSE;
sublink->useOr = FALSE;
}
else
{
/* ALL, ANY, or MULTIEXPR: generate operator list */
List *left_list = sublink->lefthand;
List *right_list = qtree->targetList;
int row_length = length(left_list);
bool needNot = false;
List *op;
char *opname;
List *elist;
/* transform lefthand expressions */
foreach(elist, left_list)
lfirst(elist) = transformExpr(pstate, lfirst(elist));
/* get the combining-operator name */
Assert(IsA(sublink->oper, A_Expr));
op = ((A_Expr *) sublink->oper)->name;
opname = strVal(llast(op));
sublink->oper = NIL;
/*
* If the expression is "<> ALL" (with unqualified opname)
* then convert it to "NOT IN". This is a hack to improve
* efficiency of expressions output by pre-7.4 Postgres.
*/
if (sublink->subLinkType == ALL_SUBLINK &&
length(op) == 1 && strcmp(opname, "<>") == 0)
{
sublink->subLinkType = ANY_SUBLINK;
opname = pstrdup("=");
op = makeList1(makeString(opname));
needNot = true;
}
/* Set operIsEquals if op is unqualified "=" */
if (length(op) == 1 && strcmp(opname, "=") == 0)
sublink->operIsEquals = TRUE;
else
sublink->operIsEquals = FALSE;
/* Set useOr if op is "<>" (possibly qualified) */
if (strcmp(opname, "<>") == 0)
sublink->useOr = TRUE;
else
sublink->useOr = FALSE;
/* Combining operators other than =/<> is dubious... */
if (length(left_list) != 1 &&
strcmp(opname, "=") != 0 && strcmp(opname, "<>") != 0)
if (row_length != 1 &&
strcmp(opname, "=") != 0 &&
strcmp(opname, "<>") != 0)
elog(ERROR, "Row comparison cannot use operator %s",
opname);
@ -474,6 +509,13 @@ transformExpr(ParseState *pstate, Node *expr)
}
if (left_list != NIL)
elog(ERROR, "Subselect has too few fields");
if (needNot)
{
expr = coerce_to_boolean(expr, "NOT");
expr = (Node *) makeBoolExpr(NOT_EXPR,
makeList1(expr));
}
}
result = (Node *) expr;
break;

View File

@ -3,7 +3,7 @@
* back to source text
*
* IDENTIFICATION
* $Header: /cvsroot/pgsql/src/backend/utils/adt/ruleutils.c,v 1.130 2003/01/08 22:54:06 momjian Exp $
* $Header: /cvsroot/pgsql/src/backend/utils/adt/ruleutils.c,v 1.131 2003/01/09 20:50:52 tgl Exp $
*
* This software is copyrighted by Jan Wieck - Hamburg.
*
@ -2660,8 +2660,16 @@ get_sublink_expr(SubLink *sublink, deparse_context *context)
break;
case ANY_SUBLINK:
oper = (OpExpr *) lfirst(sublink->oper);
appendStringInfo(buf, "%s ANY ", get_opname(oper->opno));
if (sublink->operIsEquals)
{
/* Represent it as IN */
appendStringInfo(buf, "IN ");
}
else
{
oper = (OpExpr *) lfirst(sublink->oper);
appendStringInfo(buf, "%s ANY ", get_opname(oper->opno));
}
break;
case ALL_SUBLINK:

View File

@ -37,7 +37,7 @@
* Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* $Id: catversion.h,v 1.170 2002/12/12 21:02:25 momjian Exp $
* $Id: catversion.h,v 1.171 2003/01/09 20:50:53 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -53,6 +53,6 @@
*/
/* yyyymmddN */
#define CATALOG_VERSION_NO 200212121
#define CATALOG_VERSION_NO 200301091
#endif

View File

@ -10,7 +10,7 @@
* Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* $Id: primnodes.h,v 1.75 2002/12/14 00:17:59 tgl Exp $
* $Id: primnodes.h,v 1.76 2003/01/09 20:50:53 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@ -374,7 +374,7 @@ typedef struct BoolExpr
* MULTIEXPR and EXPR require the subselect to deliver only one row.
* ALL, ANY, and MULTIEXPR require the combining operators to deliver boolean
* results. These are reduced to one result per row using OR or AND semantics
* depending on the "useor" flag. ALL and ANY combine the per-row results
* depending on the "useOr" flag. ALL and ANY combine the per-row results
* using AND and OR semantics respectively.
*
* SubLink is classed as an Expr node, but it is not actually executable;
@ -395,7 +395,11 @@ typedef struct BoolExpr
* rewriter.
*
* In EXISTS and EXPR SubLinks, both lefthand and oper are unused and are
* always NIL. useor is not significant either for these sublink types.
* always NIL. useOr is not significant either for these sublink types.
*
* The operIsEquals field is TRUE when the combining operator was written as
* "=" --- if the subLinkType is ANY_SUBLINK, this means the operation is
* equivalent to "IN". This case allows special optimizations to be used.
* ----------------
*/
typedef enum SubLinkType
@ -408,7 +412,8 @@ typedef struct SubLink
{
Expr xpr;
SubLinkType subLinkType; /* EXISTS, ALL, ANY, MULTIEXPR, EXPR */
bool useor; /* TRUE to combine column results with
bool operIsEquals; /* TRUE if combining operator is "=" */
bool useOr; /* TRUE to combine column results with
* "OR" not "AND" */
List *lefthand; /* list of outer-query expressions on the
* left */
@ -446,7 +451,7 @@ typedef struct SubPlan
Expr xpr;
/* Fields copied from original SubLink: */
SubLinkType subLinkType; /* EXISTS, ALL, ANY, MULTIEXPR, EXPR */
bool useor; /* TRUE to combine column results with
bool useOr; /* TRUE to combine column results with
* "OR" not "AND" */
List *oper; /* list of executable expressions for
* combining operators (with arguments) */