Tom Lane e6ae3b5dbf Add a concept of "placeholder" variables to the planner. These are variables
that represent some expression that we desire to compute below the top level
of the plan, and then let that value "bubble up" as though it were a plain
Var (ie, a column value).

The immediate application is to allow sub-selects to be flattened even when
they are below an outer join and have non-nullable output expressions.
Formerly we couldn't flatten because such an expression wouldn't properly
go to NULL when evaluated above the outer join.  Now, we wrap it in a
PlaceHolderVar and arrange for the actual evaluation to occur below the outer
join.  When the resulting Var bubbles up through the join, it will be set to
NULL if necessary, yielding the correct results.  This fixes a planner
limitation that's existed since 7.1.

In future we might want to use this mechanism to re-introduce some form of
Hellerstein's "expensive functions" optimization, ie place the evaluation of
an expensive function at the most suitable point in the plan tree.
2008-10-21 20:42:53 +00:00

369 lines
8.4 KiB
C

/*-------------------------------------------------------------------------
*
* tlist.c
* Target list manipulation routines
*
* Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/optimizer/util/tlist.c,v 1.83 2008/10/21 20:42:53 tgl Exp $
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#include "optimizer/tlist.h"
#include "optimizer/var.h"
#include "utils/lsyscache.h"
/*****************************************************************************
* Target list creation and searching utilities
*****************************************************************************/
/*
* tlist_member
* Finds the (first) member of the given tlist whose expression is
* equal() to the given expression. Result is NULL if no such member.
*/
TargetEntry *
tlist_member(Node *node, List *targetlist)
{
ListCell *temp;
foreach(temp, targetlist)
{
TargetEntry *tlentry = (TargetEntry *) lfirst(temp);
if (equal(node, tlentry->expr))
return tlentry;
}
return NULL;
}
/*
* tlist_member_ignore_relabel
* Same as above, except that we ignore top-level RelabelType nodes
* while checking for a match. This is needed for some scenarios
* involving binary-compatible sort operations.
*/
TargetEntry *
tlist_member_ignore_relabel(Node *node, List *targetlist)
{
ListCell *temp;
while (node && IsA(node, RelabelType))
node = (Node *) ((RelabelType *) node)->arg;
foreach(temp, targetlist)
{
TargetEntry *tlentry = (TargetEntry *) lfirst(temp);
Expr *tlexpr = tlentry->expr;
while (tlexpr && IsA(tlexpr, RelabelType))
tlexpr = ((RelabelType *) tlexpr)->arg;
if (equal(node, tlexpr))
return tlentry;
}
return NULL;
}
/*
* flatten_tlist
* Create a target list that only contains unique variables.
*
* Note that Vars with varlevelsup > 0 are not included in the output
* tlist. We expect that those will eventually be replaced with Params,
* but that probably has not happened at the time this routine is called.
*
* 'tlist' is the current target list
*
* Returns the "flattened" new target list.
*
* The result is entirely new structure sharing no nodes with the original.
* Copying the Var nodes is probably overkill, but be safe for now.
*/
List *
flatten_tlist(List *tlist)
{
List *vlist = pull_var_clause((Node *) tlist, true);
List *new_tlist;
new_tlist = add_to_flat_tlist(NIL, vlist);
list_free(vlist);
return new_tlist;
}
/*
* add_to_flat_tlist
* Add more vars to a flattened tlist (if they're not already in it)
*
* 'tlist' is the flattened tlist
* 'vars' is a list of Var and/or PlaceHolderVar nodes
*
* Returns the extended tlist.
*/
List *
add_to_flat_tlist(List *tlist, List *vars)
{
int next_resno = list_length(tlist) + 1;
ListCell *v;
foreach(v, vars)
{
Node *var = (Node *) lfirst(v);
if (!tlist_member(var, tlist))
{
TargetEntry *tle;
tle = makeTargetEntry(copyObject(var), /* copy needed?? */
next_resno++,
NULL,
false);
tlist = lappend(tlist, tle);
}
}
return tlist;
}
/*
* get_tlist_exprs
* Get just the expression subtrees of a tlist
*
* Resjunk columns are ignored unless includeJunk is true
*/
List *
get_tlist_exprs(List *tlist, bool includeJunk)
{
List *result = NIL;
ListCell *l;
foreach(l, tlist)
{
TargetEntry *tle = (TargetEntry *) lfirst(l);
if (tle->resjunk && !includeJunk)
continue;
result = lappend(result, tle->expr);
}
return result;
}
/*
* Does tlist have same output datatypes as listed in colTypes?
*
* Resjunk columns are ignored if junkOK is true; otherwise presence of
* a resjunk column will always cause a 'false' result.
*
* Note: currently no callers care about comparing typmods.
*/
bool
tlist_same_datatypes(List *tlist, List *colTypes, bool junkOK)
{
ListCell *l;
ListCell *curColType = list_head(colTypes);
foreach(l, tlist)
{
TargetEntry *tle = (TargetEntry *) lfirst(l);
if (tle->resjunk)
{
if (!junkOK)
return false;
}
else
{
if (curColType == NULL)
return false; /* tlist longer than colTypes */
if (exprType((Node *) tle->expr) != lfirst_oid(curColType))
return false;
curColType = lnext(curColType);
}
}
if (curColType != NULL)
return false; /* tlist shorter than colTypes */
return true;
}
/*
* get_sortgroupref_tle
* Find the targetlist entry matching the given SortGroupRef index,
* and return it.
*/
TargetEntry *
get_sortgroupref_tle(Index sortref, List *targetList)
{
ListCell *l;
foreach(l, targetList)
{
TargetEntry *tle = (TargetEntry *) lfirst(l);
if (tle->ressortgroupref == sortref)
return tle;
}
elog(ERROR, "ORDER/GROUP BY expression not found in targetlist");
return NULL; /* keep compiler quiet */
}
/*
* get_sortgroupclause_tle
* Find the targetlist entry matching the given SortGroupClause
* by ressortgroupref, and return it.
*/
TargetEntry *
get_sortgroupclause_tle(SortGroupClause *sgClause,
List *targetList)
{
return get_sortgroupref_tle(sgClause->tleSortGroupRef, targetList);
}
/*
* get_sortgroupclause_expr
* Find the targetlist entry matching the given SortGroupClause
* by ressortgroupref, and return its expression.
*/
Node *
get_sortgroupclause_expr(SortGroupClause *sgClause, List *targetList)
{
TargetEntry *tle = get_sortgroupclause_tle(sgClause, targetList);
return (Node *) tle->expr;
}
/*
* get_sortgrouplist_exprs
* Given a list of SortGroupClauses, build a list
* of the referenced targetlist expressions.
*/
List *
get_sortgrouplist_exprs(List *sgClauses, List *targetList)
{
List *result = NIL;
ListCell *l;
foreach(l, sgClauses)
{
SortGroupClause *sortcl = (SortGroupClause *) lfirst(l);
Node *sortexpr;
sortexpr = get_sortgroupclause_expr(sortcl, targetList);
result = lappend(result, sortexpr);
}
return result;
}
/*****************************************************************************
* 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;
}