* planner.c
* The query optimizer external interface.
*
- * Portions Copyright (c) 1996-2002, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
- * $Header: /cvsroot/pgsql/src/backend/optimizer/plan/planner.c,v 1.138 2003/01/13 18:10:53 tgl Exp $
+ * $PostgreSQL: pgsql/src/backend/optimizer/plan/planner.c,v 1.206 2006/08/02 01:59:46 joe Exp $
*
*-------------------------------------------------------------------------
*/
#include <limits.h>
#include "catalog/pg_operator.h"
-#include "catalog/pg_type.h"
+#include "executor/executor.h"
+#include "executor/nodeAgg.h"
#include "miscadmin.h"
#include "nodes/makefuncs.h"
-#ifdef OPTIMIZER_DEBUG
-#include "nodes/print.h"
-#endif
#include "optimizer/clauses.h"
#include "optimizer/cost.h"
#include "optimizer/pathnode.h"
#include "optimizer/subselect.h"
#include "optimizer/tlist.h"
#include "optimizer/var.h"
-#include "parser/analyze.h"
-#include "parser/parsetree.h"
+#ifdef OPTIMIZER_DEBUG
+#include "nodes/print.h"
+#endif
#include "parser/parse_expr.h"
#include "parser/parse_oper.h"
-#include "rewrite/rewriteManip.h"
-#include "utils/lsyscache.h"
-#include "utils/selfuncs.h"
+#include "parser/parsetree.h"
#include "utils/syscache.h"
+ParamListInfo PlannerBoundParamList = NULL; /* current boundParams */
+
+
/* Expression kind codes for preprocess_expression */
-#define EXPRKIND_TARGET 0
-#define EXPRKIND_WHERE 1
-#define EXPRKIND_HAVING 2
-
-
-static Node *pull_up_subqueries(Query *parse, Node *jtnode,
- bool below_outer_join);
-static bool is_simple_subquery(Query *subquery);
-static bool has_nullable_targetlist(Query *subquery);
-static void resolvenew_in_jointree(Node *jtnode, int varno, List *subtlist);
-static Node *preprocess_jointree(Query *parse, Node *jtnode);
-static Node *preprocess_expression(Query *parse, Node *expr, int kind);
-static void preprocess_qual_conditions(Query *parse, Node *jtnode);
-static Plan *inheritance_planner(Query *parse, List *inheritlist);
-static Plan *grouping_planner(Query *parse, double tuple_fraction);
-static bool hash_safe_grouping(Query *parse);
-static List *make_subplanTargetList(Query *parse, List *tlist,
- AttrNumber **groupColIdx);
-static Plan *make_groupsortplan(Query *parse,
- List *groupClause,
- AttrNumber *grpColIdx,
- Plan *subplan);
+#define EXPRKIND_QUAL 0
+#define EXPRKIND_TARGET 1
+#define EXPRKIND_RTFUNC 2
+#define EXPRKIND_VALUES 3
+#define EXPRKIND_LIMIT 4
+#define EXPRKIND_ININFO 5
+#define EXPRKIND_APPINFO 6
+
+
+static Node *preprocess_expression(PlannerInfo *root, Node *expr, int kind);
+static void preprocess_qual_conditions(PlannerInfo *root, Node *jtnode);
+static Plan *inheritance_planner(PlannerInfo *root);
+static Plan *grouping_planner(PlannerInfo *root, double tuple_fraction);
+static double preprocess_limit(PlannerInfo *root,
+ double tuple_fraction,
+ int64 *offset_est, int64 *count_est);
+static bool choose_hashed_grouping(PlannerInfo *root, double tuple_fraction,
+ Path *cheapest_path, Path *sorted_path,
+ double dNumGroups, AggClauseCounts *agg_counts);
+static bool hash_safe_grouping(PlannerInfo *root);
+static List *make_subplanTargetList(PlannerInfo *root, List *tlist,
+ AttrNumber **groupColIdx, bool *need_tlist_eval);
+static void locate_grouping_columns(PlannerInfo *root,
+ List *tlist,
+ List *sub_tlist,
+ AttrNumber *groupColIdx);
static List *postprocess_setop_tlist(List *new_tlist, List *orig_tlist);
*
*****************************************************************************/
Plan *
-planner(Query *parse)
+planner(Query *parse, bool isCursor, int cursorOptions,
+ ParamListInfo boundParams)
{
+ double tuple_fraction;
Plan *result_plan;
Index save_PlannerQueryLevel;
- List *save_PlannerParamVar;
+ List *save_PlannerParamList;
+ ParamListInfo save_PlannerBoundParamList;
/*
* The planner can be called recursively (an example is when
* eval_const_expressions tries to pre-evaluate an SQL function). So,
* these global state variables must be saved and restored.
*
- * These vars cannot be moved into the Query structure since their whole
- * purpose is communication across multiple sub-Queries.
+ * Query level and the param list cannot be moved into the per-query
+ * PlannerInfo structure since their whole purpose is communication across
+ * multiple sub-queries. Also, boundParams is explicitly info from outside
+ * the query, and so is likewise better handled as a global variable.
*
* Note we do NOT save and restore PlannerPlanId: it exists to assign
- * unique IDs to SubPlan nodes, and we want those IDs to be unique for
- * the life of a backend. Also, PlannerInitPlan is saved/restored in
+ * unique IDs to SubPlan nodes, and we want those IDs to be unique for the
+ * life of a backend. Also, PlannerInitPlan is saved/restored in
* subquery_planner, not here.
*/
save_PlannerQueryLevel = PlannerQueryLevel;
- save_PlannerParamVar = PlannerParamVar;
+ save_PlannerParamList = PlannerParamList;
+ save_PlannerBoundParamList = PlannerBoundParamList;
/* Initialize state for handling outer-level references and params */
PlannerQueryLevel = 0; /* will be 1 in top-level subquery_planner */
- PlannerParamVar = NIL;
+ PlannerParamList = NIL;
+ PlannerBoundParamList = boundParams;
+
+ /* Determine what fraction of the plan is likely to be scanned */
+ if (isCursor)
+ {
+ /*
+ * We have no real idea how many tuples the user will ultimately FETCH
+ * from a cursor, but it seems a good bet that he doesn't want 'em
+ * all. Optimize for 10% retrieval (you gotta better number? Should
+ * this be a SETtable parameter?)
+ */
+ tuple_fraction = 0.10;
+ }
+ else
+ {
+ /* Default assumption is we need all the tuples */
+ tuple_fraction = 0.0;
+ }
/* primary planning entry point (may recurse for subqueries) */
- result_plan = subquery_planner(parse, -1.0 /* default case */ );
+ result_plan = subquery_planner(parse, tuple_fraction, NULL);
+ /* check we popped out the right number of levels */
Assert(PlannerQueryLevel == 0);
- /* executor wants to know total number of Params used overall */
- result_plan->nParamExec = length(PlannerParamVar);
+ /*
+ * If creating a plan for a scrollable cursor, make sure it can run
+ * backwards on demand. Add a Material node at the top at need.
+ */
+ if (isCursor && (cursorOptions & CURSOR_OPT_SCROLL))
+ {
+ if (!ExecSupportsBackwardScan(result_plan))
+ result_plan = materialize_finished_plan(result_plan);
+ }
/* final cleanup of the plan */
- set_plan_references(result_plan, parse->rtable);
+ result_plan = set_plan_references(result_plan, parse->rtable);
+
+ /* executor wants to know total number of Params used overall */
+ result_plan->nParamExec = list_length(PlannerParamList);
/* restore state for outer planner, if any */
PlannerQueryLevel = save_PlannerQueryLevel;
- PlannerParamVar = save_PlannerParamVar;
+ PlannerParamList = save_PlannerParamList;
+ PlannerBoundParamList = save_PlannerBoundParamList;
return result_plan;
}
* tuple_fraction is the fraction of tuples we expect will be retrieved.
* tuple_fraction is interpreted as explained for grouping_planner, below.
*
+ * If subquery_pathkeys isn't NULL, it receives a list of pathkeys indicating
+ * the output sort ordering of the completed plan.
+ *
* Basically, this routine does the stuff that should only be done once
* per Query object. It then calls grouping_planner. At one time,
* grouping_planner could be invoked recursively on the same Query object;
*--------------------
*/
Plan *
-subquery_planner(Query *parse, double tuple_fraction)
+subquery_planner(Query *parse, double tuple_fraction,
+ List **subquery_pathkeys)
{
List *saved_initplan = PlannerInitPlan;
int saved_planid = PlannerPlanId;
+ PlannerInfo *root;
Plan *plan;
List *newHaving;
- List *lst;
+ ListCell *l;
/* Set up for a new level of subquery */
PlannerQueryLevel++;
PlannerInitPlan = NIL;
+ /* Create a PlannerInfo data structure for this subquery */
+ root = makeNode(PlannerInfo);
+ root->parse = parse;
+ root->in_info_list = NIL;
+ root->append_rel_list = NIL;
+
/*
- * Check to see if any subqueries in the rangetable can be merged into
- * this query.
+ * Look for IN clauses at the top level of WHERE, and transform them into
+ * joins. Note that this step only handles IN clauses originally at top
+ * level of WHERE; if we pull up any subqueries in the next step, their
+ * INs are processed just before pulling them up.
*/
- parse->jointree = (FromExpr *)
- pull_up_subqueries(parse, (Node *) parse->jointree, false);
+ if (parse->hasSubLinks)
+ parse->jointree->quals = pull_up_IN_clauses(root,
+ parse->jointree->quals);
/*
- * If so, we may have created opportunities to simplify the jointree.
+ * Check to see if any subqueries in the rangetable can be merged into
+ * this query.
*/
parse->jointree = (FromExpr *)
- preprocess_jointree(parse, (Node *) parse->jointree);
+ pull_up_subqueries(root, (Node *) parse->jointree, false, false);
/*
- * Detect whether any rangetable entries are RTE_JOIN kind; if not,
- * we can avoid the expense of doing flatten_join_alias_vars().
- * This must be done after we have done pull_up_subqueries, of course.
+ * Detect whether any rangetable entries are RTE_JOIN kind; if not, we can
+ * avoid the expense of doing flatten_join_alias_vars(). Also check for
+ * outer joins --- if none, we can skip reduce_outer_joins() and some
+ * other processing. This must be done after we have done
+ * pull_up_subqueries, of course.
+ *
+ * Note: if reduce_outer_joins manages to eliminate all outer joins,
+ * root->hasOuterJoins is not reset currently. This is OK since its
+ * purpose is merely to suppress unnecessary processing in simple cases.
*/
- parse->hasJoinRTEs = false;
- foreach(lst, parse->rtable)
+ root->hasJoinRTEs = false;
+ root->hasOuterJoins = false;
+ foreach(l, parse->rtable)
{
- RangeTblEntry *rte = (RangeTblEntry *) lfirst(lst);
+ RangeTblEntry *rte = (RangeTblEntry *) lfirst(l);
if (rte->rtekind == RTE_JOIN)
{
- parse->hasJoinRTEs = true;
- break;
+ root->hasJoinRTEs = true;
+ if (IS_OUTER_JOIN(rte->jointype))
+ {
+ root->hasOuterJoins = true;
+ /* Can quit scanning once we find an outer join */
+ break;
+ }
}
}
+ /*
+ * Expand any rangetable entries that are inheritance sets into "append
+ * relations". This can add entries to the rangetable, but they must be
+ * plain base relations not joins, so it's OK (and marginally more
+ * efficient) to do it after checking for join RTEs. We must do it after
+ * pulling up subqueries, else we'd fail to handle inherited tables in
+ * subqueries.
+ */
+ expand_inherited_tables(root);
+
+ /*
+ * Set hasHavingQual to remember if HAVING clause is present. Needed
+ * because preprocess_expression will reduce a constant-true condition to
+ * an empty qual list ... but "HAVING TRUE" is not a semantic no-op.
+ */
+ root->hasHavingQual = (parse->havingQual != NULL);
+
+ /* Clear this flag; might get set in distribute_qual_to_rels */
+ root->hasPseudoConstantQuals = false;
+
/*
* Do expression preprocessing on targetlist and quals.
*/
parse->targetList = (List *)
- preprocess_expression(parse, (Node *) parse->targetList,
+ preprocess_expression(root, (Node *) parse->targetList,
EXPRKIND_TARGET);
- preprocess_qual_conditions(parse, (Node *) parse->jointree);
+ preprocess_qual_conditions(root, (Node *) parse->jointree);
- parse->havingQual = preprocess_expression(parse, parse->havingQual,
- EXPRKIND_HAVING);
+ parse->havingQual = preprocess_expression(root, parse->havingQual,
+ EXPRKIND_QUAL);
- /* Also need to preprocess expressions for function RTEs */
- foreach(lst, parse->rtable)
+ parse->limitOffset = preprocess_expression(root, parse->limitOffset,
+ EXPRKIND_LIMIT);
+ parse->limitCount = preprocess_expression(root, parse->limitCount,
+ EXPRKIND_LIMIT);
+
+ root->in_info_list = (List *)
+ preprocess_expression(root, (Node *) root->in_info_list,
+ EXPRKIND_ININFO);
+ root->append_rel_list = (List *)
+ preprocess_expression(root, (Node *) root->append_rel_list,
+ EXPRKIND_APPINFO);
+
+ /* Also need to preprocess expressions for function and values RTEs */
+ foreach(l, parse->rtable)
{
- RangeTblEntry *rte = (RangeTblEntry *) lfirst(lst);
+ RangeTblEntry *rte = (RangeTblEntry *) lfirst(l);
if (rte->rtekind == RTE_FUNCTION)
- rte->funcexpr = preprocess_expression(parse, rte->funcexpr,
- EXPRKIND_TARGET);
- /* These are not targetlist items, but close enough... */
+ rte->funcexpr = preprocess_expression(root, rte->funcexpr,
+ EXPRKIND_RTFUNC);
+ else if (rte->rtekind == RTE_VALUES)
+ rte->values_lists = (List *)
+ preprocess_expression(root, (Node *) rte->values_lists,
+ EXPRKIND_VALUES);
}
/*
- * Check for ungrouped variables passed to subplans in targetlist and
- * HAVING clause (but not in WHERE or JOIN/ON clauses, since those are
- * evaluated before grouping). We can't do this any earlier because
- * we must use the preprocessed targetlist for comparisons of grouped
- * expressions.
- */
- if (parse->hasSubLinks &&
- (parse->groupClause != NIL || parse->hasAggs))
- check_subplans_for_ungrouped_vars(parse);
-
- /*
- * A HAVING clause without aggregates is equivalent to a WHERE clause
- * (except it can only refer to grouped fields). Transfer any
- * agg-free clauses of the HAVING qual into WHERE. This may seem like
- * wasting cycles to cater to stupidly-written queries, but there are
- * other reasons for doing it. Firstly, if the query contains no aggs
- * at all, then we aren't going to generate an Agg plan node, and so
- * there'll be no place to execute HAVING conditions; without this
- * transfer, we'd lose the HAVING condition entirely, which is wrong.
- * Secondly, when we push down a qual condition into a sub-query, it's
- * easiest to push the qual into HAVING always, in case it contains
- * aggs, and then let this code sort it out.
+ * In some cases we may want to transfer a HAVING clause into WHERE. We
+ * cannot do so if the HAVING clause contains aggregates (obviously) or
+ * volatile functions (since a HAVING clause is supposed to be executed
+ * only once per group). Also, it may be that the clause is so expensive
+ * to execute that we're better off doing it only once per group, despite
+ * the loss of selectivity. This is hard to estimate short of doing the
+ * entire planning process twice, so we use a heuristic: clauses
+ * containing subplans are left in HAVING. Otherwise, we move or copy the
+ * HAVING clause into WHERE, in hopes of eliminating tuples before
+ * aggregation instead of after.
+ *
+ * If the query has explicit grouping then we can simply move such a
+ * clause into WHERE; any group that fails the clause will not be in the
+ * output because none of its tuples will reach the grouping or
+ * aggregation stage. Otherwise we must have a degenerate (variable-free)
+ * HAVING clause, which we put in WHERE so that query_planner() can use it
+ * in a gating Result node, but also keep in HAVING to ensure that we
+ * don't emit a bogus aggregated row. (This could be done better, but it
+ * seems not worth optimizing.)
*
* Note that both havingQual and parse->jointree->quals are in
- * implicitly-ANDed-list form at this point, even though they are
- * declared as Node *. Also note that contain_agg_clause does not
- * recurse into sub-selects, which is exactly what we need here.
+ * implicitly-ANDed-list form at this point, even though they are declared
+ * as Node *.
*/
newHaving = NIL;
- foreach(lst, (List *) parse->havingQual)
+ foreach(l, (List *) parse->havingQual)
{
- Node *havingclause = (Node *) lfirst(lst);
+ Node *havingclause = (Node *) lfirst(l);
- if (contain_agg_clause(havingclause))
+ if (contain_agg_clause(havingclause) ||
+ contain_volatile_functions(havingclause) ||
+ contain_subplans(havingclause))
+ {
+ /* keep it in HAVING */
newHaving = lappend(newHaving, havingclause);
- else
+ }
+ else if (parse->groupClause)
+ {
+ /* move it to WHERE */
parse->jointree->quals = (Node *)
lappend((List *) parse->jointree->quals, havingclause);
+ }
+ else
+ {
+ /* put a copy in WHERE, keep it in HAVING */
+ parse->jointree->quals = (Node *)
+ lappend((List *) parse->jointree->quals,
+ copyObject(havingclause));
+ newHaving = lappend(newHaving, havingclause);
+ }
}
parse->havingQual = (Node *) newHaving;
/*
- * Do the main planning. If we have an inherited target relation,
- * that needs special processing, else go straight to
- * grouping_planner.
+ * If we have any outer joins, try to reduce them to plain inner joins.
+ * This step is most easily done after we've done expression
+ * preprocessing.
+ */
+ if (root->hasOuterJoins)
+ reduce_outer_joins(root);
+
+ /*
+ * Do the main planning. If we have an inherited target relation, that
+ * needs special processing, else go straight to grouping_planner.
*/
if (parse->resultRelation &&
- (lst = expand_inherted_rtentry(parse, parse->resultRelation, false))
- != NIL)
- plan = inheritance_planner(parse, lst);
+ rt_fetch(parse->resultRelation, parse->rtable)->inh)
+ plan = inheritance_planner(root);
else
- plan = grouping_planner(parse, tuple_fraction);
+ plan = grouping_planner(root, tuple_fraction);
/*
* If any subplans were generated, or if we're inside a subplan, build
- * initPlan, extParam and locParam lists for plan nodes.
+ * initPlan list and extParam/allParam sets for plan nodes, and attach the
+ * initPlans to the top plan node.
*/
if (PlannerPlanId != saved_planid || PlannerQueryLevel > 1)
- {
- Cost initplan_cost = 0;
+ SS_finalize_plan(plan, parse->rtable);
- /* Prepare extParam/locParam data for all nodes in tree */
- (void) SS_finalize_plan(plan, parse->rtable);
-
- /*
- * SS_finalize_plan doesn't handle initPlans, so we have to manually
- * attach them to the topmost plan node, and add their extParams to
- * the topmost node's, too.
- *
- * We also add the total_cost of each initPlan to the startup cost
- * of the top node. This is a conservative overestimate, since in
- * fact each initPlan might be executed later than plan startup, or
- * even not at all.
- */
- plan->initPlan = PlannerInitPlan;
-
- foreach(lst, plan->initPlan)
- {
- SubPlan *initplan = (SubPlan *) lfirst(lst);
-
- plan->extParam = set_unioni(plan->extParam,
- initplan->plan->extParam);
- initplan_cost += initplan->plan->total_cost;
- }
-
- plan->startup_cost += initplan_cost;
- plan->total_cost += initplan_cost;
- }
+ /* Return sort ordering info if caller wants it */
+ if (subquery_pathkeys)
+ *subquery_pathkeys = root->query_pathkeys;
/* Return to outer subquery context */
PlannerQueryLevel--;
}
/*
- * pull_up_subqueries
- * Look for subqueries in the rangetable that can be pulled up into
- * the parent query. If the subquery has no special features like
- * grouping/aggregation then we can merge it into the parent's jointree.
- *
- * below_outer_join is true if this jointree node is within the nullable
- * side of an outer join. This restricts what we can do.
- *
- * A tricky aspect of this code is that if we pull up a subquery we have
- * to replace Vars that reference the subquery's outputs throughout the
- * parent query, including quals attached to jointree nodes above the one
- * we are currently processing! We handle this by being careful not to
- * change the jointree structure while recursing: no nodes other than
- * subquery RangeTblRef entries will be replaced. Also, we can't turn
- * ResolveNew loose on the whole jointree, because it'll return a mutated
- * copy of the tree; we have to invoke it just on the quals, instead.
+ * preprocess_expression
+ * Do subquery_planner's preprocessing work for an expression,
+ * which can be a targetlist, a WHERE clause (including JOIN/ON
+ * conditions), or a HAVING clause.
*/
static Node *
-pull_up_subqueries(Query *parse, Node *jtnode, bool below_outer_join)
-{
- if (jtnode == NULL)
- return NULL;
- if (IsA(jtnode, RangeTblRef))
- {
- int varno = ((RangeTblRef *) jtnode)->rtindex;
- RangeTblEntry *rte = rt_fetch(varno, parse->rtable);
- Query *subquery = rte->subquery;
-
- /*
- * Is this a subquery RTE, and if so, is the subquery simple
- * enough to pull up? (If not, do nothing at this node.)
- *
- * If we are inside an outer join, only pull up subqueries whose
- * targetlists are nullable --- otherwise substituting their tlist
- * entries for upper Var references would do the wrong thing (the
- * results wouldn't become NULL when they're supposed to). XXX
- * This could be improved by generating pseudo-variables for such
- * expressions; we'd have to figure out how to get the pseudo-
- * variables evaluated at the right place in the modified plan
- * tree. Fix it someday.
- *
- * Note: even if the subquery itself is simple enough, we can't pull
- * it up if there is a reference to its whole tuple result.
- * Perhaps a pseudo-variable is the answer here too.
- */
- if (rte->rtekind == RTE_SUBQUERY && is_simple_subquery(subquery) &&
- (!below_outer_join || has_nullable_targetlist(subquery)) &&
- !contain_whole_tuple_var((Node *) parse, varno, 0))
- {
- int rtoffset;
- List *subtlist;
- List *rt;
-
- /*
- * First, recursively pull up the subquery's subqueries, so
- * that this routine's processing is complete for its jointree
- * and rangetable. NB: if the same subquery is referenced
- * from multiple jointree items (which can't happen normally,
- * but might after rule rewriting), then we will invoke this
- * processing multiple times on that subquery. OK because
- * nothing will happen after the first time. We do have to be
- * careful to copy everything we pull up, however, or risk
- * having chunks of structure multiply linked.
- *
- * Note: 'false' is correct here even if we are within an outer
- * join in the upper query; the lower query starts with a clean
- * slate for outer-join semantics.
- */
- subquery->jointree = (FromExpr *)
- pull_up_subqueries(subquery, (Node *) subquery->jointree,
- false);
-
- /*
- * Now make a modifiable copy of the subquery that we can run
- * OffsetVarNodes and IncrementVarSublevelsUp on.
- */
- subquery = copyObject(subquery);
-
- /*
- * Adjust level-0 varnos in subquery so that we can append its
- * rangetable to upper query's.
- */
- rtoffset = length(parse->rtable);
- OffsetVarNodes((Node *) subquery, rtoffset, 0);
-
- /*
- * Upper-level vars in subquery are now one level closer to their
- * parent than before.
- */
- IncrementVarSublevelsUp((Node *) subquery, -1, 1);
-
- /*
- * Replace all of the top query's references to the subquery's
- * outputs with copies of the adjusted subtlist items, being
- * careful not to replace any of the jointree structure.
- * (This'd be a lot cleaner if we could use
- * query_tree_mutator.)
- */
- subtlist = subquery->targetList;
- parse->targetList = (List *)
- ResolveNew((Node *) parse->targetList,
- varno, 0, subtlist, CMD_SELECT, 0);
- resolvenew_in_jointree((Node *) parse->jointree, varno, subtlist);
- Assert(parse->setOperations == NULL);
- parse->havingQual =
- ResolveNew(parse->havingQual,
- varno, 0, subtlist, CMD_SELECT, 0);
-
- foreach(rt, parse->rtable)
- {
- RangeTblEntry *rte = (RangeTblEntry *) lfirst(rt);
-
- if (rte->rtekind == RTE_JOIN)
- rte->joinaliasvars = (List *)
- ResolveNew((Node *) rte->joinaliasvars,
- varno, 0, subtlist, CMD_SELECT, 0);
- }
-
- /*
- * Now append the adjusted rtable entries to upper query. (We
- * hold off until after fixing the upper rtable entries; no
- * point in running that code on the subquery ones too.)
- */
- parse->rtable = nconc(parse->rtable, subquery->rtable);
-
- /*
- * Pull up any FOR UPDATE markers, too. (OffsetVarNodes
- * already adjusted the marker values, so just nconc the
- * list.)
- */
- parse->rowMarks = nconc(parse->rowMarks, subquery->rowMarks);
-
- /*
- * Miscellaneous housekeeping.
- */
- parse->hasSubLinks |= subquery->hasSubLinks;
- /* subquery won't be pulled up if it hasAggs, so no work there */
-
- /*
- * Return the adjusted subquery jointree to replace the
- * RangeTblRef entry in my jointree.
- */
- return (Node *) subquery->jointree;
- }
- }
- else if (IsA(jtnode, FromExpr))
- {
- FromExpr *f = (FromExpr *) jtnode;
- List *l;
-
- foreach(l, f->fromlist)
- lfirst(l) = pull_up_subqueries(parse, lfirst(l),
- below_outer_join);
- }
- else if (IsA(jtnode, JoinExpr))
- {
- JoinExpr *j = (JoinExpr *) jtnode;
-
- /* Recurse, being careful to tell myself when inside outer join */
- switch (j->jointype)
- {
- case JOIN_INNER:
- j->larg = pull_up_subqueries(parse, j->larg,
- below_outer_join);
- j->rarg = pull_up_subqueries(parse, j->rarg,
- below_outer_join);
- break;
- case JOIN_LEFT:
- j->larg = pull_up_subqueries(parse, j->larg,
- below_outer_join);
- j->rarg = pull_up_subqueries(parse, j->rarg,
- true);
- break;
- case JOIN_FULL:
- j->larg = pull_up_subqueries(parse, j->larg,
- true);
- j->rarg = pull_up_subqueries(parse, j->rarg,
- true);
- break;
- case JOIN_RIGHT:
- j->larg = pull_up_subqueries(parse, j->larg,
- true);
- j->rarg = pull_up_subqueries(parse, j->rarg,
- below_outer_join);
- break;
- case JOIN_UNION:
-
- /*
- * This is where we fail if upper levels of planner
- * haven't rewritten UNION JOIN as an Append ...
- */
- elog(ERROR, "UNION JOIN is not implemented yet");
- break;
- default:
- elog(ERROR, "pull_up_subqueries: unexpected join type %d",
- j->jointype);
- break;
- }
- }
- else
- elog(ERROR, "pull_up_subqueries: unexpected node type %d",
- nodeTag(jtnode));
- return jtnode;
-}
-
-/*
- * is_simple_subquery
- * Check a subquery in the range table to see if it's simple enough
- * to pull up into the parent query.
- */
-static bool
-is_simple_subquery(Query *subquery)
+preprocess_expression(PlannerInfo *root, Node *expr, int kind)
{
/*
- * Let's just make sure it's a valid subselect ...
+ * Fall out quickly if expression is empty. This occurs often enough to
+ * be worth checking. Note that null->null is the correct conversion for
+ * implicit-AND result format, too.
*/
- if (!IsA(subquery, Query) ||
- subquery->commandType != CMD_SELECT ||
- subquery->resultRelation != 0 ||
- subquery->into != NULL ||
- subquery->isPortal)
- elog(ERROR, "is_simple_subquery: subquery is bogus");
+ if (expr == NULL)
+ return NULL;
/*
- * Can't currently pull up a query with setops. Maybe after querytree
- * redesign...
+ * If the query has any join RTEs, replace join alias variables with
+ * base-relation variables. We must do this before sublink processing,
+ * else sublinks expanded out from join aliases wouldn't get processed.
+ * We can skip it in VALUES lists, however, since they can't contain
+ * any Vars at all.
*/
- if (subquery->setOperations)
- return false;
+ if (root->hasJoinRTEs && kind != EXPRKIND_VALUES)
+ expr = flatten_join_alias_vars(root, expr);
/*
- * Can't pull up a subquery involving grouping, aggregation, sorting,
- * or limiting.
+ * Simplify constant expressions.
+ *
+ * Note: this also flattens nested AND and OR expressions into N-argument
+ * form. All processing of a qual expression after this point must be
+ * careful to maintain AND/OR flatness --- that is, do not generate a tree
+ * with AND directly under AND, nor OR directly under OR.
+ *
+ * Because this is a relatively expensive process, we skip it when the
+ * query is trivial, such as "SELECT 2+2;" or "INSERT ... VALUES()". The
+ * expression will only be evaluated once anyway, so no point in
+ * pre-simplifying; we can't execute it any faster than the executor can,
+ * and we will waste cycles copying the tree. Notice however that we
+ * still must do it for quals (to get AND/OR flatness); and if we are in a
+ * subquery we should not assume it will be done only once.
+ *
+ * For VALUES lists we never do this at all, again on the grounds that
+ * we should optimize for one-time evaluation.
*/
- if (subquery->hasAggs ||
- subquery->groupClause ||
- subquery->havingQual ||
- subquery->sortClause ||
- subquery->distinctClause ||
- subquery->limitOffset ||
- subquery->limitCount)
- return false;
+ if (kind != EXPRKIND_VALUES &&
+ (root->parse->jointree->fromlist != NIL ||
+ kind == EXPRKIND_QUAL ||
+ PlannerQueryLevel > 1))
+ expr = eval_const_expressions(expr);
/*
- * Don't pull up a subquery that has any set-returning functions in
- * its targetlist. Otherwise we might well wind up inserting
- * set-returning functions into places where they mustn't go, such as
- * quals of higher queries.
+ * If it's a qual or havingQual, canonicalize it.
*/
- if (expression_returns_set((Node *) subquery->targetList))
- return false;
+ if (kind == EXPRKIND_QUAL)
+ {
+ expr = (Node *) canonicalize_qual((Expr *) expr);
- /*
- * Don't pull up a subquery that has any sublinks in its targetlist,
- * either. As of PG 7.3 this creates problems because the pulled-up
- * expressions may go into join alias lists, and the sublinks would
- * not get fixed because we do flatten_join_alias_vars() too late.
- * Eventually we should do a complete flatten_join_alias_vars as the
- * first step of preprocess_expression, and then we could probably
- * support this. (BUT: it might be a bad idea anyway, due to possibly
- * causing multiple evaluations of an expensive sublink.)
- */
- if (subquery->hasSubLinks &&
- contain_subplans((Node *) subquery->targetList))
- return false;
+#ifdef OPTIMIZER_DEBUG
+ printf("After canonicalize_qual()\n");
+ pprint(expr);
+#endif
+ }
+
+ /* Expand SubLinks to SubPlans */
+ if (root->parse->hasSubLinks)
+ expr = SS_process_sublinks(expr, (kind == EXPRKIND_QUAL));
/*
- * Hack: don't try to pull up a subquery with an empty jointree.
- * query_planner() will correctly generate a Result plan for a
- * jointree that's totally empty, but I don't think the right things
- * happen if an empty FromExpr appears lower down in a jointree. Not
- * worth working hard on this, just to collapse SubqueryScan/Result
- * into Result...
+ * XXX do not insert anything here unless you have grokked the comments in
+ * SS_replace_correlation_vars ...
*/
- if (subquery->jointree->fromlist == NIL)
- return false;
-
- return true;
-}
-
-/*
- * has_nullable_targetlist
- * Check a subquery in the range table to see if all the non-junk
- * targetlist items are simple variables (and, hence, will correctly
- * go to NULL when examined above the point of an outer join).
- *
- * A possible future extension is to accept strict functions of simple
- * variables, eg, "x + 1".
- */
-static bool
-has_nullable_targetlist(Query *subquery)
-{
- List *l;
-
- foreach(l, subquery->targetList)
- {
- TargetEntry *tle = (TargetEntry *) lfirst(l);
- /* ignore resjunk columns */
- if (tle->resdom->resjunk)
- continue;
+ /* Replace uplevel vars with Param nodes (this IS possible in VALUES) */
+ if (PlannerQueryLevel > 1)
+ expr = SS_replace_correlation_vars(expr);
- /* Okay if tlist item is a simple Var */
- if (tle->expr && IsA(tle->expr, Var))
- continue;
+ /*
+ * If it's a qual or havingQual, convert it to implicit-AND format. (We
+ * don't want to do this before eval_const_expressions, since the latter
+ * would be unable to simplify a top-level AND correctly. Also,
+ * SS_process_sublinks expects explicit-AND format.)
+ */
+ if (kind == EXPRKIND_QUAL)
+ expr = (Node *) make_ands_implicit((Expr *) expr);
- return false;
- }
- return true;
+ return expr;
}
/*
- * Helper routine for pull_up_subqueries: do ResolveNew on every expression
- * in the jointree, without changing the jointree structure itself. Ugly,
- * but there's no other way...
+ * preprocess_qual_conditions
+ * Recursively scan the query's jointree and do subquery_planner's
+ * preprocessing work on each qual condition found therein.
*/
static void
-resolvenew_in_jointree(Node *jtnode, int varno, List *subtlist)
+preprocess_qual_conditions(PlannerInfo *root, Node *jtnode)
{
if (jtnode == NULL)
return;
else if (IsA(jtnode, FromExpr))
{
FromExpr *f = (FromExpr *) jtnode;
- List *l;
+ ListCell *l;
foreach(l, f->fromlist)
- resolvenew_in_jointree(lfirst(l), varno, subtlist);
- f->quals = ResolveNew(f->quals,
- varno, 0, subtlist, CMD_SELECT, 0);
+ preprocess_qual_conditions(root, lfirst(l));
+
+ f->quals = preprocess_expression(root, f->quals, EXPRKIND_QUAL);
}
else if (IsA(jtnode, JoinExpr))
{
JoinExpr *j = (JoinExpr *) jtnode;
- resolvenew_in_jointree(j->larg, varno, subtlist);
- resolvenew_in_jointree(j->rarg, varno, subtlist);
- j->quals = ResolveNew(j->quals,
- varno, 0, subtlist, CMD_SELECT, 0);
+ preprocess_qual_conditions(root, j->larg);
+ preprocess_qual_conditions(root, j->rarg);
- /*
- * We don't bother to update the colvars list, since it won't be
- * used again ...
- */
+ j->quals = preprocess_expression(root, j->quals, EXPRKIND_QUAL);
}
else
- elog(ERROR, "resolvenew_in_jointree: unexpected node type %d",
- nodeTag(jtnode));
+ elog(ERROR, "unrecognized node type: %d",
+ (int) nodeTag(jtnode));
}
/*
- * preprocess_jointree
- * Attempt to simplify a query's jointree.
+ * inheritance_planner
+ * Generate a plan in the case where the result relation is an
+ * inheritance set.
*
- * If we succeed in pulling up a subquery then we might form a jointree
- * in which a FromExpr is a direct child of another FromExpr. In that
- * case we can consider collapsing the two FromExprs into one. This is
- * an optional conversion, since the planner will work correctly either
- * way. But we may find a better plan (at the cost of more planning time)
- * if we merge the two nodes.
+ * We have to handle this case differently from cases where a source relation
+ * is an inheritance set. Source inheritance is expanded at the bottom of the
+ * plan tree (see allpaths.c), but target inheritance has to be expanded at
+ * the top. The reason is that for UPDATE, each target relation needs a
+ * different targetlist matching its own column set. Also, for both UPDATE
+ * and DELETE, the executor needs the Append plan node at the top, else it
+ * can't keep track of which table is the current target table. Fortunately,
+ * the UPDATE/DELETE target can never be the nullable side of an outer join,
+ * so it's OK to generate the plan this way.
*
- * NOTE: don't try to do this in the same jointree scan that does subquery
- * pullup! Since we're changing the jointree structure here, that wouldn't
- * work reliably --- see comments for pull_up_subqueries().
+ * Returns a query plan.
*/
-static Node *
-preprocess_jointree(Query *parse, Node *jtnode)
+static Plan *
+inheritance_planner(PlannerInfo *root)
{
- if (jtnode == NULL)
- return NULL;
- if (IsA(jtnode, RangeTblRef))
- {
- /* nothing to do here... */
- }
- else if (IsA(jtnode, FromExpr))
- {
- FromExpr *f = (FromExpr *) jtnode;
- List *newlist = NIL;
- List *l;
+ Query *parse = root->parse;
+ int parentRTindex = parse->resultRelation;
+ List *subplans = NIL;
+ List *tlist = NIL;
+ PlannerInfo subroot;
+ ListCell *l;
- foreach(l, f->fromlist)
- {
- Node *child = (Node *) lfirst(l);
+ subroot.parse = NULL; /* catch it if no matches in loop */
- /* Recursively simplify the child... */
- child = preprocess_jointree(parse, child);
- /* Now, is it a FromExpr? */
- if (child && IsA(child, FromExpr))
- {
- /*
- * Yes, so do we want to merge it into parent? Always do
- * so if child has just one element (since that doesn't
- * make the parent's list any longer). Otherwise we have
- * to be careful about the increase in planning time
- * caused by combining the two join search spaces into
- * one. Our heuristic is to merge if the merge will
- * produce a join list no longer than GEQO_RELS/2.
- * (Perhaps need an additional user parameter?)
- */
- FromExpr *subf = (FromExpr *) child;
- int childlen = length(subf->fromlist);
- int myothers = length(newlist) + length(lnext(l));
+ parse->resultRelations = NIL;
- if (childlen <= 1 || (childlen + myothers) <= geqo_rels / 2)
- {
- newlist = nconc(newlist, subf->fromlist);
- f->quals = make_and_qual(subf->quals, f->quals);
- }
- else
- newlist = lappend(newlist, child);
- }
- else
- newlist = lappend(newlist, child);
- }
- f->fromlist = newlist;
- }
- else if (IsA(jtnode, JoinExpr))
+ foreach(l, root->append_rel_list)
{
- JoinExpr *j = (JoinExpr *) jtnode;
+ AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l);
+ Plan *subplan;
+
+ /* append_rel_list contains all append rels; ignore others */
+ if (appinfo->parent_relid != parentRTindex)
+ continue;
+
+ /* Build target-relations list for the executor */
+ parse->resultRelations = lappend_int(parse->resultRelations,
+ appinfo->child_relid);
+
+ /*
+ * Generate modified query with this rel as target. We have to be
+ * prepared to translate varnos in in_info_list as well as in the
+ * Query proper.
+ */
+ memcpy(&subroot, root, sizeof(PlannerInfo));
+ subroot.parse = (Query *)
+ adjust_appendrel_attrs((Node *) parse,
+ appinfo);
+ subroot.in_info_list = (List *)
+ adjust_appendrel_attrs((Node *) root->in_info_list,
+ appinfo);
+ /* There shouldn't be any OJ info to translate, as yet */
+ Assert(subroot.oj_info_list == NIL);
+
+ /* Generate plan */
+ subplan = grouping_planner(&subroot, 0.0 /* retrieve all tuples */ );
+
+ subplans = lappend(subplans, subplan);
- /* Can't usefully change the JoinExpr, but recurse on children */
- j->larg = preprocess_jointree(parse, j->larg);
- j->rarg = preprocess_jointree(parse, j->rarg);
+ /* Save preprocessed tlist from first rel for use in Append */
+ if (tlist == NIL)
+ tlist = subplan->targetlist;
}
- else
- elog(ERROR, "preprocess_jointree: unexpected node type %d",
- nodeTag(jtnode));
- return jtnode;
-}
-/*
- * preprocess_expression
- * Do subquery_planner's preprocessing work for an expression,
- * which can be a targetlist, a WHERE clause (including JOIN/ON
- * conditions), or a HAVING clause.
- */
-static Node *
-preprocess_expression(Query *parse, Node *expr, int kind)
-{
/*
- * Simplify constant expressions.
+ * Planning might have modified the rangetable, due to changes of the
+ * Query structures inside subquery RTEs. We have to ensure that this
+ * gets propagated back to the master copy. But can't do this until we
+ * are done planning, because all the calls to grouping_planner need
+ * virgin sub-Queries to work from. (We are effectively assuming that
+ * sub-Queries will get planned identically each time, or at least that
+ * the impacts on their rangetables will be the same each time.)
*
- * Note that at this point quals have not yet been converted to
- * implicit-AND form, so we can apply eval_const_expressions directly.
+ * XXX should clean this up someday
*/
- expr = eval_const_expressions(expr);
+ parse->rtable = subroot.parse->rtable;
- /*
- * If it's a qual or havingQual, canonicalize it, and convert it to
- * implicit-AND format.
- *
- * XXX Is there any value in re-applying eval_const_expressions after
- * canonicalize_qual?
- */
- if (kind != EXPRKIND_TARGET)
- {
- expr = (Node *) canonicalize_qual((Expr *) expr, true);
-
-#ifdef OPTIMIZER_DEBUG
- printf("After canonicalize_qual()\n");
- pprint(expr);
-#endif
- }
-
- /* Expand SubLinks to SubPlans */
- if (parse->hasSubLinks)
- expr = SS_process_sublinks(expr, (kind != EXPRKIND_TARGET));
-
- /* Replace uplevel vars with Param nodes */
- if (PlannerQueryLevel > 1)
- expr = SS_replace_correlation_vars(expr);
-
- /*
- * If the query has any join RTEs, try to replace join alias variables
- * with base-relation variables, to allow quals to be pushed down. We
- * must do this after sublink processing, since it does not recurse
- * into sublinks.
- */
- if (parse->hasJoinRTEs)
- expr = flatten_join_alias_vars(expr, parse->rtable, false);
-
- return expr;
-}
-
-/*
- * preprocess_qual_conditions
- * Recursively scan the query's jointree and do subquery_planner's
- * preprocessing work on each qual condition found therein.
- */
-static void
-preprocess_qual_conditions(Query *parse, Node *jtnode)
-{
- if (jtnode == NULL)
- return;
- if (IsA(jtnode, RangeTblRef))
- {
- /* nothing to do here */
- }
- else if (IsA(jtnode, FromExpr))
- {
- FromExpr *f = (FromExpr *) jtnode;
- List *l;
-
- foreach(l, f->fromlist)
- preprocess_qual_conditions(parse, lfirst(l));
-
- f->quals = preprocess_expression(parse, f->quals, EXPRKIND_WHERE);
- }
- else if (IsA(jtnode, JoinExpr))
- {
- JoinExpr *j = (JoinExpr *) jtnode;
-
- preprocess_qual_conditions(parse, j->larg);
- preprocess_qual_conditions(parse, j->rarg);
-
- j->quals = preprocess_expression(parse, j->quals, EXPRKIND_WHERE);
- }
- else
- elog(ERROR, "preprocess_qual_conditions: unexpected node type %d",
- nodeTag(jtnode));
-}
-
-/*--------------------
- * inheritance_planner
- * Generate a plan in the case where the result relation is an
- * inheritance set.
- *
- * We have to handle this case differently from cases where a source
- * relation is an inheritance set. Source inheritance is expanded at
- * the bottom of the plan tree (see allpaths.c), but target inheritance
- * has to be expanded at the top. The reason is that for UPDATE, each
- * target relation needs a different targetlist matching its own column
- * set. (This is not so critical for DELETE, but for simplicity we treat
- * inherited DELETE the same way.) Fortunately, the UPDATE/DELETE target
- * can never be the nullable side of an outer join, so it's OK to generate
- * the plan this way.
- *
- * parse is the querytree produced by the parser & rewriter.
- * inheritlist is an integer list of RT indexes for the result relation set.
- *
- * Returns a query plan.
- *--------------------
- */
-static Plan *
-inheritance_planner(Query *parse, List *inheritlist)
-{
- int parentRTindex = parse->resultRelation;
- Oid parentOID = getrelid(parentRTindex, parse->rtable);
- List *subplans = NIL;
- List *tlist = NIL;
- List *l;
-
- foreach(l, inheritlist)
- {
- int childRTindex = lfirsti(l);
- Oid childOID = getrelid(childRTindex, parse->rtable);
- Query *subquery;
- Plan *subplan;
-
- /* Generate modified query with this rel as target */
- subquery = (Query *) adjust_inherited_attrs((Node *) parse,
- parentRTindex, parentOID,
- childRTindex, childOID);
- /* Generate plan */
- subplan = grouping_planner(subquery, 0.0 /* retrieve all tuples */ );
- subplans = lappend(subplans, subplan);
- /* Save preprocessed tlist from first rel for use in Append */
- if (tlist == NIL)
- tlist = subplan->targetlist;
- }
-
- /* Save the target-relations list for the executor, too */
- parse->resultRelations = inheritlist;
+ /* Mark result as unordered (probably unnecessary) */
+ root->query_pathkeys = NIL;
return (Plan *) make_append(subplans, true, tlist);
}
* This primarily means adding top-level processing to the basic
* query plan produced by query_planner.
*
- * parse is the querytree produced by the parser & rewriter.
* tuple_fraction is the fraction of tuples we expect will be retrieved
*
* tuple_fraction is interpreted as follows:
- * < 0: determine fraction by inspection of query (normal case)
- * 0: expect all tuples to be retrieved
+ * 0: expect all tuples to be retrieved (normal case)
* 0 < tuple_fraction < 1: expect the given fraction of tuples available
* from the plan to be retrieved
* tuple_fraction >= 1: tuple_fraction is the absolute number of tuples
* expected to be retrieved (ie, a LIMIT specification)
- * The normal case is to pass -1, but some callers pass values >= 0 to
- * override this routine's determination of the appropriate fraction.
*
- * Returns a query plan.
+ * Returns a query plan. Also, root->query_pathkeys is returned as the
+ * actual output ordering of the plan (in pathkey format).
*--------------------
*/
static Plan *
-grouping_planner(Query *parse, double tuple_fraction)
+grouping_planner(PlannerInfo *root, double tuple_fraction)
{
+ Query *parse = root->parse;
List *tlist = parse->targetList;
+ int64 offset_est = 0;
+ int64 count_est = 0;
Plan *result_plan;
List *current_pathkeys;
List *sort_pathkeys;
+ double dNumGroups = 0;
+
+ /* Tweak caller-supplied tuple_fraction if have LIMIT/OFFSET */
+ if (parse->limitCount || parse->limitOffset)
+ tuple_fraction = preprocess_limit(root, tuple_fraction,
+ &offset_est, &count_est);
if (parse->setOperations)
{
+ List *set_sortclauses;
+
/*
- * Construct the plan for set operations. The result will not
- * need any work except perhaps a top-level sort and/or LIMIT.
+ * If there's a top-level ORDER BY, assume we have to fetch all the
+ * tuples. This might seem too simplistic given all the hackery below
+ * to possibly avoid the sort ... but a nonzero tuple_fraction is only
+ * of use to plan_set_operations() when the setop is UNION ALL, and
+ * the result of UNION ALL is always unsorted.
*/
- result_plan = plan_set_operations(parse);
+ if (parse->sortClause)
+ tuple_fraction = 0.0;
/*
- * We should not need to call preprocess_targetlist, since we must
- * be in a SELECT query node. Instead, use the targetlist
- * returned by plan_set_operations (since this tells whether it
- * returned any resjunk columns!), and transfer any sort key
- * information from the original tlist.
+ * Construct the plan for set operations. The result will not need
+ * any work except perhaps a top-level sort and/or LIMIT.
+ */
+ result_plan = plan_set_operations(root, tuple_fraction,
+ &set_sortclauses);
+
+ /*
+ * Calculate pathkeys representing the sort order (if any) of the set
+ * operation's result. We have to do this before overwriting the sort
+ * key information...
+ */
+ current_pathkeys = make_pathkeys_for_sortclauses(set_sortclauses,
+ result_plan->targetlist);
+ current_pathkeys = canonicalize_pathkeys(root, current_pathkeys);
+
+ /*
+ * We should not need to call preprocess_targetlist, since we must be
+ * in a SELECT query node. Instead, use the targetlist returned by
+ * plan_set_operations (since this tells whether it returned any
+ * resjunk columns!), and transfer any sort key information from the
+ * original tlist.
*/
Assert(parse->commandType == CMD_SELECT);
tlist = postprocess_setop_tlist(result_plan->targetlist, tlist);
/*
- * Can't handle FOR UPDATE here (parser should have checked
+ * Can't handle FOR UPDATE/SHARE here (parser should have checked
* already, but let's make sure).
*/
if (parse->rowMarks)
- elog(ERROR, "SELECT FOR UPDATE is not allowed with UNION/INTERSECT/EXCEPT");
-
- /*
- * We set current_pathkeys NIL indicating we do not know sort
- * order. This is correct when the top set operation is UNION
- * ALL, since the appended-together results are unsorted even if
- * the subplans were sorted. For other set operations we could be
- * smarter --- room for future improvement!
- */
- current_pathkeys = NIL;
+ ereport(ERROR,
+ (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+ errmsg("SELECT FOR UPDATE/SHARE is not allowed with UNION/INTERSECT/EXCEPT")));
/*
- * Calculate pathkeys that represent ordering requirements
+ * Calculate pathkeys that represent result ordering requirements
*/
sort_pathkeys = make_pathkeys_for_sortclauses(parse->sortClause,
tlist);
- sort_pathkeys = canonicalize_pathkeys(parse, sort_pathkeys);
+ sort_pathkeys = canonicalize_pathkeys(root, sort_pathkeys);
}
else
{
List *sub_tlist;
List *group_pathkeys;
AttrNumber *groupColIdx = NULL;
+ bool need_tlist_eval = true;
QualCost tlist_cost;
- double sub_tuple_fraction;
Path *cheapest_path;
Path *sorted_path;
- double dNumGroups = 0;
+ Path *best_path;
long numGroups = 0;
- int numAggs = 0;
- int numGroupCols = length(parse->groupClause);
+ AggClauseCounts agg_counts;
+ int numGroupCols = list_length(parse->groupClause);
bool use_hashed_grouping = false;
- /* Preprocess targetlist in case we are inside an INSERT/UPDATE. */
- tlist = preprocess_targetlist(tlist,
- parse->commandType,
- parse->resultRelation,
- parse->rtable);
+ MemSet(&agg_counts, 0, sizeof(AggClauseCounts));
- /*
- * Add TID targets for rels selected FOR UPDATE (should this be
- * done in preprocess_targetlist?). The executor uses the TID to
- * know which rows to lock, much as for UPDATE or DELETE.
- */
- if (parse->rowMarks)
- {
- List *l;
-
- /*
- * We've got trouble if the FOR UPDATE appears inside
- * grouping, since grouping renders a reference to individual
- * tuple CTIDs invalid. This is also checked at parse time,
- * but that's insufficient because of rule substitution, query
- * pullup, etc.
- */
- CheckSelectForUpdate(parse);
-
- /*
- * Currently the executor only supports FOR UPDATE at top
- * level
- */
- if (PlannerQueryLevel > 1)
- elog(ERROR, "SELECT FOR UPDATE is not allowed in subselects");
-
- foreach(l, parse->rowMarks)
- {
- Index rti = lfirsti(l);
- char *resname;
- Resdom *resdom;
- Var *var;
- TargetEntry *ctid;
-
- resname = (char *) palloc(32);
- snprintf(resname, 32, "ctid%u", rti);
- resdom = makeResdom(length(tlist) + 1,
- TIDOID,
- -1,
- resname,
- true);
-
- var = makeVar(rti,
- SelfItemPointerAttributeNumber,
- TIDOID,
- -1,
- 0);
-
- ctid = makeTargetEntry(resdom, (Expr *) var);
- tlist = lappend(tlist, ctid);
- }
- }
+ /* Preprocess targetlist */
+ tlist = preprocess_targetlist(root, tlist);
/*
- * Generate appropriate target list for subplan; may be different
- * from tlist if grouping or aggregation is needed.
+ * Generate appropriate target list for subplan; may be different from
+ * tlist if grouping or aggregation is needed.
*/
- sub_tlist = make_subplanTargetList(parse, tlist, &groupColIdx);
+ sub_tlist = make_subplanTargetList(root, tlist,
+ &groupColIdx, &need_tlist_eval);
/*
- * Calculate pathkeys that represent grouping/ordering
- * requirements
+ * Calculate pathkeys that represent grouping/ordering requirements.
+ * Stash them in PlannerInfo so that query_planner can canonicalize
+ * them.
*/
- group_pathkeys = make_pathkeys_for_sortclauses(parse->groupClause,
- tlist);
- sort_pathkeys = make_pathkeys_for_sortclauses(parse->sortClause,
- tlist);
+ root->group_pathkeys =
+ make_pathkeys_for_sortclauses(parse->groupClause, tlist);
+ root->sort_pathkeys =
+ make_pathkeys_for_sortclauses(parse->sortClause, tlist);
/*
* Will need actual number of aggregates for estimating costs.
- * Also, it's possible that optimization has eliminated all
- * aggregates, and we may as well check for that here.
+ *
+ * Note: we do not attempt to detect duplicate aggregates here; a
+ * somewhat-overestimated count is okay for our present purposes.
+ *
+ * Note: think not that we can turn off hasAggs if we find no aggs. It
+ * is possible for constant-expression simplification to remove all
+ * explicit references to aggs, but we still have to follow the
+ * aggregate semantics (eg, producing only one output row).
*/
if (parse->hasAggs)
{
- numAggs = length(pull_agg_clause((Node *) tlist)) +
- length(pull_agg_clause(parse->havingQual));
- if (numAggs == 0)
- parse->hasAggs = false;
+ count_agg_clauses((Node *) tlist, &agg_counts);
+ count_agg_clauses(parse->havingQual, &agg_counts);
}
/*
* Figure out whether we need a sorted result from query_planner.
*
- * If we have a GROUP BY clause, then we want a result sorted
- * properly for grouping. Otherwise, if there is an ORDER BY
- * clause, we want to sort by the ORDER BY clause. (Note: if we
- * have both, and ORDER BY is a superset of GROUP BY, it would be
- * tempting to request sort by ORDER BY --- but that might just
- * leave us failing to exploit an available sort order at all.
- * Needs more thought...)
+ * If we have a GROUP BY clause, then we want a result sorted properly
+ * for grouping. Otherwise, if there is an ORDER BY clause, we want
+ * to sort by the ORDER BY clause. (Note: if we have both, and ORDER
+ * BY is a superset of GROUP BY, it would be tempting to request sort
+ * by ORDER BY --- but that might just leave us failing to exploit an
+ * available sort order at all. Needs more thought...)
*/
if (parse->groupClause)
- parse->query_pathkeys = group_pathkeys;
+ root->query_pathkeys = root->group_pathkeys;
else if (parse->sortClause)
- parse->query_pathkeys = sort_pathkeys;
+ root->query_pathkeys = root->sort_pathkeys;
else
- parse->query_pathkeys = NIL;
+ root->query_pathkeys = NIL;
/*
- * Figure out whether we expect to retrieve all the tuples that
- * the plan can generate, or to stop early due to outside factors
- * such as a cursor. If the caller passed a value >= 0, believe
- * that value, else do our own examination of the query context.
+ * Generate the best unsorted and presorted paths for this Query (but
+ * note there may not be any presorted path). query_planner will also
+ * estimate the number of groups in the query, and canonicalize all
+ * the pathkeys.
*/
- if (tuple_fraction < 0.0)
+ query_planner(root, sub_tlist, tuple_fraction,
+ &cheapest_path, &sorted_path, &dNumGroups);
+
+ group_pathkeys = root->group_pathkeys;
+ sort_pathkeys = root->sort_pathkeys;
+
+ /*
+ * If grouping, decide whether we want to use hashed grouping.
+ */
+ if (parse->groupClause)
{
- /* Initial assumption is we need all the tuples */
- tuple_fraction = 0.0;
+ use_hashed_grouping =
+ choose_hashed_grouping(root, tuple_fraction,
+ cheapest_path, sorted_path,
+ dNumGroups, &agg_counts);
- /*
- * Check for retrieve-into-portal, ie DECLARE CURSOR.
- *
- * We have no real idea how many tuples the user will ultimately
- * FETCH from a cursor, but it seems a good bet that he
- * doesn't want 'em all. Optimize for 10% retrieval (you
- * gotta better number? Should this be a SETtable parameter?)
- */
- if (parse->isPortal)
- tuple_fraction = 0.10;
+ /* Also convert # groups to long int --- but 'ware overflow! */
+ numGroups = (long) Min(dNumGroups, (double) LONG_MAX);
}
/*
- * Adjust tuple_fraction if we see that we are going to apply
- * limiting/grouping/aggregation/etc. This is not overridable by
- * the caller, since it reflects plan actions that this routine
- * will certainly take, not assumptions about context.
+ * Select the best path. If we are doing hashed grouping, we will
+ * always read all the input tuples, so use the cheapest-total path.
+ * Otherwise, trust query_planner's decision about which to use.
+ */
+ if (use_hashed_grouping || !sorted_path)
+ best_path = cheapest_path;
+ else
+ best_path = sorted_path;
+
+ /*
+ * Check to see if it's possible to optimize MIN/MAX aggregates. If
+ * so, we will forget all the work we did so far to choose a "regular"
+ * path ... but we had to do it anyway to be able to tell which way is
+ * cheaper.
*/
- if (parse->limitCount != NULL)
+ result_plan = optimize_minmax_aggregates(root,
+ tlist,
+ best_path);
+ if (result_plan != NULL)
{
/*
- * A LIMIT clause limits the absolute number of tuples
- * returned. However, if it's not a constant LIMIT then we
- * have to punt; for lack of a better idea, assume 10% of the
- * plan's result is wanted.
+ * optimize_minmax_aggregates generated the full plan, with the
+ * right tlist, and it has no sort order.
*/
- double limit_fraction = 0.0;
+ current_pathkeys = NIL;
+ }
+ else
+ {
+ /*
+ * Normal case --- create a plan according to query_planner's
+ * results.
+ */
+ result_plan = create_plan(root, best_path);
+ current_pathkeys = best_path->pathkeys;
- if (IsA(parse->limitCount, Const))
+ /*
+ * create_plan() returns a plan with just a "flat" tlist of
+ * required Vars. Usually we need to insert the sub_tlist as the
+ * tlist of the top plan node. However, we can skip that if we
+ * determined that whatever query_planner chose to return will be
+ * good enough.
+ */
+ if (need_tlist_eval)
{
- Const *limitc = (Const *) parse->limitCount;
- int32 count = DatumGetInt32(limitc->constvalue);
-
/*
- * A NULL-constant LIMIT represents "LIMIT ALL", which we
- * treat the same as no limit (ie, expect to retrieve all
- * the tuples).
+ * If the top-level plan node is one that cannot do expression
+ * evaluation, we must insert a Result node to project the
+ * desired tlist.
*/
- if (!limitc->constisnull && count > 0)
+ if (!is_projection_capable_plan(result_plan))
{
- limit_fraction = (double) count;
- /* We must also consider the OFFSET, if present */
- if (parse->limitOffset != NULL)
- {
- if (IsA(parse->limitOffset, Const))
- {
- int32 offset;
-
- limitc = (Const *) parse->limitOffset;
- offset = DatumGetInt32(limitc->constvalue);
- if (!limitc->constisnull && offset > 0)
- limit_fraction += (double) offset;
- }
- else
- {
- /* OFFSET is an expression ... punt ... */
- limit_fraction = 0.10;
- }
- }
+ result_plan = (Plan *) make_result(sub_tlist, NULL,
+ result_plan);
}
+ else
+ {
+ /*
+ * Otherwise, just replace the subplan's flat tlist with
+ * the desired tlist.
+ */
+ result_plan->targetlist = sub_tlist;
+ }
+
+ /*
+ * Also, account for the cost of evaluation of the sub_tlist.
+ *
+ * Up to now, we have only been dealing with "flat" tlists,
+ * containing just Vars. So their evaluation cost is zero
+ * according to the model used by cost_qual_eval() (or if you
+ * prefer, the cost is factored into cpu_tuple_cost). Thus we
+ * can avoid accounting for tlist cost throughout
+ * query_planner() and subroutines. But now we've inserted a
+ * tlist that might contain actual operators, sub-selects, etc
+ * --- so we'd better account for its cost.
+ *
+ * Below this point, any tlist eval cost for added-on nodes
+ * should be accounted for as we create those nodes.
+ * Presently, of the node types we can add on, only Agg and
+ * Group project new tlists (the rest just copy their input
+ * tuples) --- so make_agg() and make_group() are responsible
+ * for computing the added cost.
+ */
+ cost_qual_eval(&tlist_cost, sub_tlist);
+ result_plan->startup_cost += tlist_cost.startup;
+ result_plan->total_cost += tlist_cost.startup +
+ tlist_cost.per_tuple * result_plan->plan_rows;
}
else
{
- /* LIMIT is an expression ... punt ... */
- limit_fraction = 0.10;
+ /*
+ * Since we're using query_planner's tlist and not the one
+ * make_subplanTargetList calculated, we have to refigure any
+ * grouping-column indexes make_subplanTargetList computed.
+ */
+ locate_grouping_columns(root, tlist, result_plan->targetlist,
+ groupColIdx);
}
- if (limit_fraction > 0.0)
+ /*
+ * Insert AGG or GROUP node if needed, plus an explicit sort step
+ * if necessary.
+ *
+ * HAVING clause, if any, becomes qual of the Agg or Group node.
+ */
+ if (use_hashed_grouping)
{
- /*
- * If we have absolute limits from both caller and LIMIT,
- * use the smaller value; if one is fractional and the
- * other absolute, treat the fraction as a fraction of the
- * absolute value; else we can multiply the two fractions
- * together.
- */
- if (tuple_fraction >= 1.0)
+ /* Hashed aggregate plan --- no sort needed */
+ result_plan = (Plan *) make_agg(root,
+ tlist,
+ (List *) parse->havingQual,
+ AGG_HASHED,
+ numGroupCols,
+ groupColIdx,
+ numGroups,
+ agg_counts.numAggs,
+ result_plan);
+ /* Hashed aggregation produces randomly-ordered results */
+ current_pathkeys = NIL;
+ }
+ else if (parse->hasAggs)
+ {
+ /* Plain aggregate plan --- sort if needed */
+ AggStrategy aggstrategy;
+
+ if (parse->groupClause)
{
- if (limit_fraction >= 1.0)
- {
- /* both absolute */
- tuple_fraction = Min(tuple_fraction, limit_fraction);
- }
- else
+ if (!pathkeys_contained_in(group_pathkeys,
+ current_pathkeys))
{
- /* caller absolute, limit fractional */
- tuple_fraction *= limit_fraction;
- if (tuple_fraction < 1.0)
- tuple_fraction = 1.0;
+ result_plan = (Plan *)
+ make_sort_from_groupcols(root,
+ parse->groupClause,
+ groupColIdx,
+ result_plan);
+ current_pathkeys = group_pathkeys;
}
+ aggstrategy = AGG_SORTED;
+
+ /*
+ * The AGG node will not change the sort ordering of its
+ * groups, so current_pathkeys describes the result too.
+ */
}
- else if (tuple_fraction > 0.0)
+ else
{
- if (limit_fraction >= 1.0)
- {
- /* caller fractional, limit absolute */
- tuple_fraction *= limit_fraction;
- if (tuple_fraction < 1.0)
- tuple_fraction = 1.0;
- }
- else
- {
- /* both fractional */
- tuple_fraction *= limit_fraction;
- }
+ aggstrategy = AGG_PLAIN;
+ /* Result will be only one row anyway; no sort order */
+ current_pathkeys = NIL;
}
- else
+
+ result_plan = (Plan *) make_agg(root,
+ tlist,
+ (List *) parse->havingQual,
+ aggstrategy,
+ numGroupCols,
+ groupColIdx,
+ numGroups,
+ agg_counts.numAggs,
+ result_plan);
+ }
+ else if (parse->groupClause)
+ {
+ /*
+ * GROUP BY without aggregation, so insert a group node (plus
+ * the appropriate sort node, if necessary).
+ *
+ * Add an explicit sort if we couldn't make the path come out
+ * the way the GROUP node needs it.
+ */
+ if (!pathkeys_contained_in(group_pathkeys, current_pathkeys))
{
- /* no info from caller, just use limit */
- tuple_fraction = limit_fraction;
+ result_plan = (Plan *)
+ make_sort_from_groupcols(root,
+ parse->groupClause,
+ groupColIdx,
+ result_plan);
+ current_pathkeys = group_pathkeys;
}
+
+ result_plan = (Plan *) make_group(root,
+ tlist,
+ (List *) parse->havingQual,
+ numGroupCols,
+ groupColIdx,
+ dNumGroups,
+ result_plan);
+ /* The Group node won't change sort ordering */
}
+ else if (root->hasHavingQual)
+ {
+ /*
+ * No aggregates, and no GROUP BY, but we have a HAVING qual.
+ * This is a degenerate case in which we are supposed to emit
+ * either 0 or 1 row depending on whether HAVING succeeds.
+ * Furthermore, there cannot be any variables in either HAVING
+ * or the targetlist, so we actually do not need the FROM
+ * table at all! We can just throw away the plan-so-far and
+ * generate a Result node. This is a sufficiently unusual
+ * corner case that it's not worth contorting the structure of
+ * this routine to avoid having to generate the plan in the
+ * first place.
+ */
+ result_plan = (Plan *) make_result(tlist,
+ parse->havingQual,
+ NULL);
+ }
+ } /* end of non-minmax-aggregate case */
+ } /* end of if (setOperations) */
+
+ /*
+ * If we were not able to make the plan come out in the right order, add
+ * an explicit sort step.
+ */
+ if (parse->sortClause)
+ {
+ if (!pathkeys_contained_in(sort_pathkeys, current_pathkeys))
+ {
+ result_plan = (Plan *)
+ make_sort_from_sortclauses(root,
+ parse->sortClause,
+ result_plan);
+ current_pathkeys = sort_pathkeys;
}
+ }
+
+ /*
+ * If there is a DISTINCT clause, add the UNIQUE node.
+ */
+ if (parse->distinctClause)
+ {
+ result_plan = (Plan *) make_unique(result_plan, parse->distinctClause);
/*
- * With grouping or aggregation, the tuple fraction to pass to
- * query_planner() may be different from what it is at top level.
+ * If there was grouping or aggregation, leave plan_rows as-is (ie,
+ * assume the result was already mostly unique). If not, use the
+ * number of distinct-groups calculated by query_planner.
*/
- sub_tuple_fraction = tuple_fraction;
+ if (!parse->groupClause && !root->hasHavingQual && !parse->hasAggs)
+ result_plan->plan_rows = dNumGroups;
+ }
- if (parse->groupClause)
- {
- /*
- * In GROUP BY mode, we have the little problem that we don't
- * really know how many input tuples will be needed to make a
- * group, so we can't translate an output LIMIT count into an
- * input count. For lack of a better idea, assume 25% of the
- * input data will be processed if there is any output limit.
- * However, if the caller gave us a fraction rather than an
- * absolute count, we can keep using that fraction (which
- * amounts to assuming that all the groups are about the same
- * size).
- */
- if (sub_tuple_fraction >= 1.0)
- sub_tuple_fraction = 0.25;
+ /*
+ * Finally, if there is a LIMIT/OFFSET clause, add the LIMIT node.
+ */
+ if (parse->limitCount || parse->limitOffset)
+ {
+ result_plan = (Plan *) make_limit(result_plan,
+ parse->limitOffset,
+ parse->limitCount,
+ offset_est,
+ count_est);
+ }
- /*
- * If both GROUP BY and ORDER BY are specified, we will need
- * two levels of sort --- and, therefore, certainly need to
- * read all the input tuples --- unless ORDER BY is a subset
- * of GROUP BY. (We have not yet canonicalized the pathkeys,
- * so must use the slower noncanonical comparison method.)
- */
- if (parse->groupClause && parse->sortClause &&
- !noncanonical_pathkeys_contained_in(sort_pathkeys,
- group_pathkeys))
- sub_tuple_fraction = 0.0;
- }
- else if (parse->hasAggs)
- {
- /*
- * Ungrouped aggregate will certainly want all the input
- * tuples.
- */
- sub_tuple_fraction = 0.0;
- }
- else if (parse->distinctClause)
- {
- /*
- * SELECT DISTINCT, like GROUP, will absorb an unpredictable
- * number of input tuples per output tuple. Handle the same
- * way.
- */
- if (sub_tuple_fraction >= 1.0)
- sub_tuple_fraction = 0.25;
- }
+ /*
+ * Return the actual output ordering in query_pathkeys for possible use by
+ * an outer query level.
+ */
+ root->query_pathkeys = current_pathkeys;
- /*
- * Generate the best unsorted and presorted paths for this Query
- * (but note there may not be any presorted path).
- */
- query_planner(parse, sub_tlist, sub_tuple_fraction,
- &cheapest_path, &sorted_path);
+ return result_plan;
+}
- /*
- * We couldn't canonicalize group_pathkeys and sort_pathkeys before
- * running query_planner(), so do it now.
- */
- group_pathkeys = canonicalize_pathkeys(parse, group_pathkeys);
- sort_pathkeys = canonicalize_pathkeys(parse, sort_pathkeys);
+/*
+ * preprocess_limit - do pre-estimation for LIMIT and/or OFFSET clauses
+ *
+ * We try to estimate the values of the LIMIT/OFFSET clauses, and pass the
+ * results back in *count_est and *offset_est. These variables are set to
+ * 0 if the corresponding clause is not present, and -1 if it's present
+ * but we couldn't estimate the value for it. (The "0" convention is OK
+ * for OFFSET but a little bit bogus for LIMIT: effectively we estimate
+ * LIMIT 0 as though it were LIMIT 1. But this is in line with the planner's
+ * usual practice of never estimating less than one row.) These values will
+ * be passed to make_limit, which see if you change this code.
+ *
+ * The return value is the suitably adjusted tuple_fraction to use for
+ * planning the query. This adjustment is not overridable, since it reflects
+ * plan actions that grouping_planner() will certainly take, not assumptions
+ * about context.
+ */
+static double
+preprocess_limit(PlannerInfo *root, double tuple_fraction,
+ int64 *offset_est, int64 *count_est)
+{
+ Query *parse = root->parse;
+ Node *est;
+ double limit_fraction;
- /*
- * Consider whether we might want to use hashed grouping.
- */
- if (parse->groupClause)
- {
- /*
- * Always estimate the number of groups. We can't do this until
- * after running query_planner(), either.
- */
- dNumGroups = estimate_num_groups(parse,
- parse->groupClause,
- cheapest_path->parent->rows);
- /* Also want it as a long int --- but 'ware overflow! */
- numGroups = (long) Min(dNumGroups, (double) LONG_MAX);
+ /* Should not be called unless LIMIT or OFFSET */
+ Assert(parse->limitCount || parse->limitOffset);
- /*
- * Check can't-do-it conditions, including whether the grouping
- * operators are hashjoinable.
- *
- * Executor doesn't support hashed aggregation with DISTINCT
- * aggregates. (Doing so would imply storing *all* the input
- * values in the hash table, which seems like a certain loser.)
- */
- if (!enable_hashagg || !hash_safe_grouping(parse))
- use_hashed_grouping = false;
- else if (parse->hasAggs &&
- (contain_distinct_agg_clause((Node *) tlist) ||
- contain_distinct_agg_clause(parse->havingQual)))
- use_hashed_grouping = false;
+ /*
+ * Try to obtain the clause values. We use estimate_expression_value
+ * primarily because it can sometimes do something useful with Params.
+ */
+ if (parse->limitCount)
+ {
+ est = estimate_expression_value(parse->limitCount);
+ if (est && IsA(est, Const))
+ {
+ if (((Const *) est)->constisnull)
+ {
+ /* NULL indicates LIMIT ALL, ie, no limit */
+ *count_est = 0; /* treat as not present */
+ }
else
{
- /*
- * Use hashed grouping if (a) we think we can fit the
- * hashtable into SortMem, *and* (b) the estimated cost
- * is no more than doing it the other way. While avoiding
- * the need for sorted input is usually a win, the fact
- * that the output won't be sorted may be a loss; so we
- * need to do an actual cost comparison.
- *
- * In most cases we have no good way to estimate the size of
- * the transition value needed by an aggregate; arbitrarily
- * assume it is 100 bytes. Also set the overhead per hashtable
- * entry at 64 bytes.
- */
- int hashentrysize = cheapest_path->parent->width + 64 +
- numAggs * 100;
-
- if (hashentrysize * dNumGroups <= SortMem * 1024L)
- {
- /*
- * Okay, do the cost comparison. We need to consider
- * cheapest_path + hashagg [+ final sort]
- * versus either
- * cheapest_path [+ sort] + group or agg [+ final sort]
- * or
- * presorted_path + group or agg [+ final sort]
- * where brackets indicate a step that may not be needed.
- * We assume query_planner() will have returned a
- * presorted path only if it's a winner compared to
- * cheapest_path for this purpose.
- *
- * These path variables are dummies that just hold cost
- * fields; we don't make actual Paths for these steps.
- */
- Path hashed_p;
- Path sorted_p;
-
- cost_agg(&hashed_p, parse,
- AGG_HASHED, numAggs,
- numGroupCols, dNumGroups,
- cheapest_path->startup_cost,
- cheapest_path->total_cost,
- cheapest_path->parent->rows);
- /* Result of hashed agg is always unsorted */
- if (sort_pathkeys)
- cost_sort(&hashed_p, parse, sort_pathkeys,
- hashed_p.total_cost,
- dNumGroups,
- cheapest_path->parent->width);
-
- if (sorted_path)
- {
- sorted_p.startup_cost = sorted_path->startup_cost;
- sorted_p.total_cost = sorted_path->total_cost;
- current_pathkeys = sorted_path->pathkeys;
- }
- else
- {
- sorted_p.startup_cost = cheapest_path->startup_cost;
- sorted_p.total_cost = cheapest_path->total_cost;
- current_pathkeys = cheapest_path->pathkeys;
- }
- if (!pathkeys_contained_in(group_pathkeys,
- current_pathkeys))
- {
- cost_sort(&sorted_p, parse, group_pathkeys,
- sorted_p.total_cost,
- cheapest_path->parent->rows,
- cheapest_path->parent->width);
- current_pathkeys = group_pathkeys;
- }
- if (parse->hasAggs)
- cost_agg(&sorted_p, parse,
- AGG_SORTED, numAggs,
- numGroupCols, dNumGroups,
- sorted_p.startup_cost,
- sorted_p.total_cost,
- cheapest_path->parent->rows);
- else
- cost_group(&sorted_p, parse,
- numGroupCols, dNumGroups,
- sorted_p.startup_cost,
- sorted_p.total_cost,
- cheapest_path->parent->rows);
- /* The Agg or Group node will preserve ordering */
- if (sort_pathkeys &&
- !pathkeys_contained_in(sort_pathkeys,
- current_pathkeys))
- {
- cost_sort(&sorted_p, parse, sort_pathkeys,
- sorted_p.total_cost,
- dNumGroups,
- cheapest_path->parent->width);
- }
-
- /*
- * Now make the decision using the top-level tuple
- * fraction. First we have to convert an absolute
- * count (LIMIT) into fractional form.
- */
- if (tuple_fraction >= 1.0)
- tuple_fraction /= dNumGroups;
+ *count_est = DatumGetInt64(((Const *) est)->constvalue);
+ if (*count_est <= 0)
+ *count_est = 1; /* force to at least 1 */
+ }
+ }
+ else
+ *count_est = -1; /* can't estimate */
+ }
+ else
+ *count_est = 0; /* not present */
- if (compare_fractional_path_costs(&hashed_p, &sorted_p,
- tuple_fraction) <= 0)
- {
- /* Hashed is cheaper, so use it */
- use_hashed_grouping = true;
- }
- }
+ if (parse->limitOffset)
+ {
+ est = estimate_expression_value(parse->limitOffset);
+ if (est && IsA(est, Const))
+ {
+ if (((Const *) est)->constisnull)
+ {
+ /* Treat NULL as no offset; the executor will too */
+ *offset_est = 0; /* treat as not present */
+ }
+ else
+ {
+ *offset_est = DatumGetInt64(((Const *) est)->constvalue);
+ if (*offset_est < 0)
+ *offset_est = 0; /* less than 0 is same as 0 */
}
}
+ else
+ *offset_est = -1; /* can't estimate */
+ }
+ else
+ *offset_est = 0; /* not present */
+ if (*count_est != 0)
+ {
/*
- * Select the best path and create a plan to execute it.
- *
- * If we are doing hashed grouping, we will always read all the
- * input tuples, so use the cheapest-total path. Otherwise,
- * trust query_planner's decision about which to use.
+ * A LIMIT clause limits the absolute number of tuples returned.
+ * However, if it's not a constant LIMIT then we have to guess; for
+ * lack of a better idea, assume 10% of the plan's result is wanted.
*/
- if (sorted_path && !use_hashed_grouping)
+ if (*count_est < 0 || *offset_est < 0)
{
- result_plan = create_plan(parse, sorted_path);
- current_pathkeys = sorted_path->pathkeys;
+ /* LIMIT or OFFSET is an expression ... punt ... */
+ limit_fraction = 0.10;
}
else
{
- result_plan = create_plan(parse, cheapest_path);
- current_pathkeys = cheapest_path->pathkeys;
+ /* LIMIT (plus OFFSET, if any) is max number of tuples needed */
+ limit_fraction = (double) *count_est + (double) *offset_est;
}
/*
- * create_plan() returns a plan with just a "flat" tlist of required
- * Vars. We want to insert the sub_tlist as the tlist of the top
- * plan node. If the top-level plan node is one that cannot do
- * expression evaluation, we must insert a Result node to project the
- * desired tlist.
- * Currently, the only plan node we might see here that falls into
- * that category is Append.
+ * If we have absolute limits from both caller and LIMIT, use the
+ * smaller value; likewise if they are both fractional. If one is
+ * fractional and the other absolute, we can't easily determine which
+ * is smaller, but we use the heuristic that the absolute will usually
+ * be smaller.
*/
- if (IsA(result_plan, Append))
+ if (tuple_fraction >= 1.0)
{
- result_plan = (Plan *) make_result(sub_tlist, NULL, result_plan);
+ if (limit_fraction >= 1.0)
+ {
+ /* both absolute */
+ tuple_fraction = Min(tuple_fraction, limit_fraction);
+ }
+ else
+ {
+ /* caller absolute, limit fractional; use caller's value */
+ }
+ }
+ else if (tuple_fraction > 0.0)
+ {
+ if (limit_fraction >= 1.0)
+ {
+ /* caller fractional, limit absolute; use limit */
+ tuple_fraction = limit_fraction;
+ }
+ else
+ {
+ /* both fractional */
+ tuple_fraction = Min(tuple_fraction, limit_fraction);
+ }
}
else
{
- /*
- * Otherwise, just replace the flat tlist with the desired tlist.
- */
- result_plan->targetlist = sub_tlist;
+ /* no info from caller, just use limit */
+ tuple_fraction = limit_fraction;
}
+ }
+ else if (*offset_est != 0 && tuple_fraction > 0.0)
+ {
/*
- * Also, account for the cost of evaluation of the sub_tlist.
+ * We have an OFFSET but no LIMIT. This acts entirely differently
+ * from the LIMIT case: here, we need to increase rather than decrease
+ * the caller's tuple_fraction, because the OFFSET acts to cause more
+ * tuples to be fetched instead of fewer. This only matters if we got
+ * a tuple_fraction > 0, however.
*
- * Up to now, we have only been dealing with "flat" tlists, containing
- * just Vars. So their evaluation cost is zero according to the
- * model used by cost_qual_eval() (or if you prefer, the cost is
- * factored into cpu_tuple_cost). Thus we can avoid accounting for
- * tlist cost throughout query_planner() and subroutines.
- * But now we've inserted a tlist that might contain actual operators,
- * sub-selects, etc --- so we'd better account for its cost.
- *
- * Below this point, any tlist eval cost for added-on nodes should
- * be accounted for as we create those nodes. Presently, of the
- * node types we can add on, only Agg and Group project new tlists
- * (the rest just copy their input tuples) --- so make_agg() and
- * make_group() are responsible for computing the added cost.
+ * As above, use 10% if OFFSET is present but unestimatable.
*/
- cost_qual_eval(&tlist_cost, sub_tlist);
- result_plan->startup_cost += tlist_cost.startup;
- result_plan->total_cost += tlist_cost.startup +
- tlist_cost.per_tuple * result_plan->plan_rows;
+ if (*offset_est < 0)
+ limit_fraction = 0.10;
+ else
+ limit_fraction = (double) *offset_est;
/*
- * Insert AGG or GROUP node if needed, plus an explicit sort step
- * if necessary.
- *
- * HAVING clause, if any, becomes qual of the Agg node
+ * If we have absolute counts from both caller and OFFSET, add them
+ * together; likewise if they are both fractional. If one is
+ * fractional and the other absolute, we want to take the larger, and
+ * we heuristically assume that's the fractional one.
*/
- if (use_hashed_grouping)
+ if (tuple_fraction >= 1.0)
{
- /* Hashed aggregate plan --- no sort needed */
- result_plan = (Plan *) make_agg(parse,
- tlist,
- (List *) parse->havingQual,
- AGG_HASHED,
- numGroupCols,
- groupColIdx,
- numGroups,
- numAggs,
- result_plan);
- /* Hashed aggregation produces randomly-ordered results */
- current_pathkeys = NIL;
- }
- else if (parse->hasAggs)
- {
- /* Plain aggregate plan --- sort if needed */
- AggStrategy aggstrategy;
-
- if (parse->groupClause)
+ if (limit_fraction >= 1.0)
{
- if (!pathkeys_contained_in(group_pathkeys, current_pathkeys))
- {
- result_plan = make_groupsortplan(parse,
- parse->groupClause,
- groupColIdx,
- result_plan);
- current_pathkeys = group_pathkeys;
- }
- aggstrategy = AGG_SORTED;
- /*
- * The AGG node will not change the sort ordering of its
- * groups, so current_pathkeys describes the result too.
- */
+ /* both absolute, so add them together */
+ tuple_fraction += limit_fraction;
}
else
{
- aggstrategy = AGG_PLAIN;
- /* Result will be only one row anyway; no sort order */
- current_pathkeys = NIL;
+ /* caller absolute, limit fractional; use limit */
+ tuple_fraction = limit_fraction;
}
-
- result_plan = (Plan *) make_agg(parse,
- tlist,
- (List *) parse->havingQual,
- aggstrategy,
- numGroupCols,
- groupColIdx,
- numGroups,
- numAggs,
- result_plan);
}
else
{
- /*
- * If there are no Aggs, we shouldn't have any HAVING qual anymore
- */
- Assert(parse->havingQual == NULL);
-
- /*
- * If we have a GROUP BY clause, insert a group node (plus the
- * appropriate sort node, if necessary).
- */
- if (parse->groupClause)
+ if (limit_fraction >= 1.0)
{
- /*
- * Add an explicit sort if we couldn't make the path come out
- * the way the GROUP node needs it.
- */
- if (!pathkeys_contained_in(group_pathkeys, current_pathkeys))
- {
- result_plan = make_groupsortplan(parse,
- parse->groupClause,
- groupColIdx,
- result_plan);
- current_pathkeys = group_pathkeys;
- }
-
- result_plan = (Plan *) make_group(parse,
- tlist,
- numGroupCols,
- groupColIdx,
- dNumGroups,
- result_plan);
- /* The Group node won't change sort ordering */
+ /* caller fractional, limit absolute; use caller's value */
+ }
+ else
+ {
+ /* both fractional, so add them together */
+ tuple_fraction += limit_fraction;
+ if (tuple_fraction >= 1.0)
+ tuple_fraction = 0.0; /* assume fetch all */
}
}
- } /* end of if (setOperations) */
+ }
+
+ return tuple_fraction;
+}
+
+/*
+ * choose_hashed_grouping - should we use hashed grouping?
+ */
+static bool
+choose_hashed_grouping(PlannerInfo *root, double tuple_fraction,
+ Path *cheapest_path, Path *sorted_path,
+ double dNumGroups, AggClauseCounts *agg_counts)
+{
+ int numGroupCols = list_length(root->parse->groupClause);
+ double cheapest_path_rows;
+ int cheapest_path_width;
+ Size hashentrysize;
+ List *current_pathkeys;
+ Path hashed_p;
+ Path sorted_p;
/*
- * If we were not able to make the plan come out in the right order,
- * add an explicit sort step.
+ * Check can't-do-it conditions, including whether the grouping operators
+ * are hashjoinable.
+ *
+ * Executor doesn't support hashed aggregation with DISTINCT aggregates.
+ * (Doing so would imply storing *all* the input values in the hash table,
+ * which seems like a certain loser.)
*/
- if (parse->sortClause)
+ if (!enable_hashagg)
+ return false;
+ if (agg_counts->numDistinctAggs != 0)
+ return false;
+ if (!hash_safe_grouping(root))
+ return false;
+
+ /*
+ * Don't do it if it doesn't look like the hashtable will fit into
+ * work_mem.
+ *
+ * Beware here of the possibility that cheapest_path->parent is NULL. This
+ * could happen if user does something silly like SELECT 'foo' GROUP BY 1;
+ */
+ if (cheapest_path->parent)
{
- if (!pathkeys_contained_in(sort_pathkeys, current_pathkeys))
- result_plan = make_sortplan(parse, tlist, result_plan,
- parse->sortClause);
+ cheapest_path_rows = cheapest_path->parent->rows;
+ cheapest_path_width = cheapest_path->parent->width;
+ }
+ else
+ {
+ cheapest_path_rows = 1; /* assume non-set result */
+ cheapest_path_width = 100; /* arbitrary */
}
+ /* Estimate per-hash-entry space at tuple width... */
+ hashentrysize = MAXALIGN(cheapest_path_width) + MAXALIGN(sizeof(MinimalTupleData));
+ /* plus space for pass-by-ref transition values... */
+ hashentrysize += agg_counts->transitionSpace;
+ /* plus the per-hash-entry overhead */
+ hashentrysize += hash_agg_entry_size(agg_counts->numAggs);
+
+ if (hashentrysize * dNumGroups > work_mem * 1024L)
+ return false;
+
/*
- * If there is a DISTINCT clause, add the UNIQUE node.
+ * See if the estimated cost is no more than doing it the other way. While
+ * avoiding the need for sorted input is usually a win, the fact that the
+ * output won't be sorted may be a loss; so we need to do an actual cost
+ * comparison.
+ *
+ * We need to consider cheapest_path + hashagg [+ final sort] versus
+ * either cheapest_path [+ sort] + group or agg [+ final sort] or
+ * presorted_path + group or agg [+ final sort] where brackets indicate a
+ * step that may not be needed. We assume query_planner() will have
+ * returned a presorted path only if it's a winner compared to
+ * cheapest_path for this purpose.
+ *
+ * These path variables are dummies that just hold cost fields; we don't
+ * make actual Paths for these steps.
*/
- if (parse->distinctClause)
+ cost_agg(&hashed_p, root, AGG_HASHED, agg_counts->numAggs,
+ numGroupCols, dNumGroups,
+ cheapest_path->startup_cost, cheapest_path->total_cost,
+ cheapest_path_rows);
+ /* Result of hashed agg is always unsorted */
+ if (root->sort_pathkeys)
+ cost_sort(&hashed_p, root, root->sort_pathkeys, hashed_p.total_cost,
+ dNumGroups, cheapest_path_width);
+
+ if (sorted_path)
{
- result_plan = (Plan *) make_unique(tlist, result_plan,
- parse->distinctClause);
- /*
- * If there was grouping or aggregation, leave plan_rows as-is
- * (ie, assume the result was already mostly unique). If not,
- * it's reasonable to assume the UNIQUE filter has effects
- * comparable to GROUP BY.
- */
- if (!parse->groupClause && !parse->hasAggs)
- result_plan->plan_rows = estimate_num_groups(parse,
- parse->distinctClause,
- result_plan->plan_rows);
+ sorted_p.startup_cost = sorted_path->startup_cost;
+ sorted_p.total_cost = sorted_path->total_cost;
+ current_pathkeys = sorted_path->pathkeys;
+ }
+ else
+ {
+ sorted_p.startup_cost = cheapest_path->startup_cost;
+ sorted_p.total_cost = cheapest_path->total_cost;
+ current_pathkeys = cheapest_path->pathkeys;
+ }
+ if (!pathkeys_contained_in(root->group_pathkeys, current_pathkeys))
+ {
+ cost_sort(&sorted_p, root, root->group_pathkeys, sorted_p.total_cost,
+ cheapest_path_rows, cheapest_path_width);
+ current_pathkeys = root->group_pathkeys;
}
+ if (root->parse->hasAggs)
+ cost_agg(&sorted_p, root, AGG_SORTED, agg_counts->numAggs,
+ numGroupCols, dNumGroups,
+ sorted_p.startup_cost, sorted_p.total_cost,
+ cheapest_path_rows);
+ else
+ cost_group(&sorted_p, root, numGroupCols, dNumGroups,
+ sorted_p.startup_cost, sorted_p.total_cost,
+ cheapest_path_rows);
+ /* The Agg or Group node will preserve ordering */
+ if (root->sort_pathkeys &&
+ !pathkeys_contained_in(root->sort_pathkeys, current_pathkeys))
+ cost_sort(&sorted_p, root, root->sort_pathkeys, sorted_p.total_cost,
+ dNumGroups, cheapest_path_width);
+
/*
- * Finally, if there is a LIMIT/OFFSET clause, add the LIMIT node.
+ * Now make the decision using the top-level tuple fraction. First we
+ * have to convert an absolute count (LIMIT) into fractional form.
*/
- if (parse->limitOffset || parse->limitCount)
+ if (tuple_fraction >= 1.0)
+ tuple_fraction /= dNumGroups;
+
+ if (compare_fractional_path_costs(&hashed_p, &sorted_p,
+ tuple_fraction) < 0)
{
- result_plan = (Plan *) make_limit(tlist, result_plan,
- parse->limitOffset,
- parse->limitCount);
+ /* Hashed is cheaper, so use it */
+ return true;
}
-
- return result_plan;
+ return false;
}
/*
* is marked hashjoinable.
*/
static bool
-hash_safe_grouping(Query *parse)
+hash_safe_grouping(PlannerInfo *root)
{
- List *gl;
+ ListCell *gl;
- foreach(gl, parse->groupClause)
+ foreach(gl, root->parse->groupClause)
{
GroupClause *grpcl = (GroupClause *) lfirst(gl);
- TargetEntry *tle = get_sortgroupclause_tle(grpcl, parse->targetList);
+ TargetEntry *tle = get_sortgroupclause_tle(grpcl,
+ root->parse->targetList);
Operator optup;
bool oprcanhash;
- optup = equality_oper(tle->resdom->restype, false);
+ optup = equality_oper(exprType((Node *) tle->expr), true);
+ if (!optup)
+ return false;
oprcanhash = ((Form_pg_operator) GETSTRUCT(optup))->oprcanhash;
ReleaseSysCache(optup);
if (!oprcanhash)
* make_subplanTargetList
* Generate appropriate target list when grouping is required.
*
- * When grouping_planner inserts Aggregate or Group plan nodes above
- * the result of query_planner, we typically want to pass a different
+ * When grouping_planner inserts Aggregate, Group, or Result plan nodes
+ * above the result of query_planner, we typically want to pass a different
* target list to query_planner than the outer plan nodes should have.
* This routine generates the correct target list for the subplan.
*
* The initial target list passed from the parser already contains entries
* for all ORDER BY and GROUP BY expressions, but it will not have entries
* for variables used only in HAVING clauses; so we need to add those
- * variables to the subplan target list. Also, if we are doing either
- * grouping or aggregation, we flatten all expressions except GROUP BY items
- * into their component variables; the other expressions will be computed by
- * the inserted nodes rather than by the subplan. For example,
- * given a query like
+ * variables to the subplan target list. Also, we flatten all expressions
+ * except GROUP BY items into their component variables; the other expressions
+ * will be computed by the inserted nodes rather than by the subplan.
+ * For example, given a query like
* SELECT a+b,SUM(c+d) FROM table GROUP BY a+b;
* we want to pass this targetlist to the subplan:
* a,b,c,d,a+b
* the extra computation to recompute a+b at the outer level; see
* replace_vars_with_subplan_refs() in setrefs.c.)
*
- * 'parse' is the query being processed.
+ * If we are grouping or aggregating, *and* there are no non-Var grouping
+ * expressions, then the returned tlist is effectively dummy; we do not
+ * need to force it to be evaluated, because all the Vars it contains
+ * should be present in the output of query_planner anyway.
+ *
* 'tlist' is the query's target list.
* 'groupColIdx' receives an array of column numbers for the GROUP BY
- * expressions (if there are any) in the subplan's target list.
+ * expressions (if there are any) in the subplan's target list.
+ * 'need_tlist_eval' is set true if we really need to evaluate the
+ * result tlist.
*
* The result is the targetlist to be passed to the subplan.
*---------------
*/
static List *
-make_subplanTargetList(Query *parse,
+make_subplanTargetList(PlannerInfo *root,
List *tlist,
- AttrNumber **groupColIdx)
+ AttrNumber **groupColIdx,
+ bool *need_tlist_eval)
{
+ Query *parse = root->parse;
List *sub_tlist;
List *extravars;
int numCols;
*groupColIdx = NULL;
/*
- * If we're not grouping or aggregating, nothing to do here;
+ * If we're not grouping or aggregating, there's nothing to do here;
* query_planner should receive the unmodified target list.
*/
- if (!parse->hasAggs && !parse->groupClause && !parse->havingQual)
+ if (!parse->hasAggs && !parse->groupClause && !root->hasHavingQual)
+ {
+ *need_tlist_eval = true;
return tlist;
+ }
/*
* Otherwise, start with a "flattened" tlist (having just the vars
- * mentioned in the targetlist and HAVING qual --- but not upper-
- * level Vars; they will be replaced by Params later on).
+ * mentioned in the targetlist and HAVING qual --- but not upper- level
+ * Vars; they will be replaced by Params later on).
*/
sub_tlist = flatten_tlist(tlist);
extravars = pull_var_clause(parse->havingQual, false);
sub_tlist = add_to_flat_tlist(sub_tlist, extravars);
- freeList(extravars);
+ list_free(extravars);
+ *need_tlist_eval = false; /* only eval if not flat tlist */
/*
* If grouping, create sub_tlist entries for all GROUP BY expressions
- * (GROUP BY items that are simple Vars should be in the list
- * already), and make an array showing where the group columns are in
- * the sub_tlist.
+ * (GROUP BY items that are simple Vars should be in the list already),
+ * and make an array showing where the group columns are in the sub_tlist.
*/
- numCols = length(parse->groupClause);
+ numCols = list_length(parse->groupClause);
if (numCols > 0)
{
int keyno = 0;
AttrNumber *grpColIdx;
- List *gl;
+ ListCell *gl;
grpColIdx = (AttrNumber *) palloc(sizeof(AttrNumber) * numCols);
*groupColIdx = grpColIdx;
GroupClause *grpcl = (GroupClause *) lfirst(gl);
Node *groupexpr = get_sortgroupclause_expr(grpcl, tlist);
TargetEntry *te = NULL;
- List *sl;
+ ListCell *sl;
/* Find or make a matching sub_tlist entry */
foreach(sl, sub_tlist)
}
if (!sl)
{
- te = makeTargetEntry(makeResdom(length(sub_tlist) + 1,
- exprType(groupexpr),
- exprTypmod(groupexpr),
- NULL,
- false),
- (Expr *) groupexpr);
+ te = makeTargetEntry((Expr *) groupexpr,
+ list_length(sub_tlist) + 1,
+ NULL,
+ false);
sub_tlist = lappend(sub_tlist, te);
+ *need_tlist_eval = true; /* it's not flat anymore */
}
/* and save its resno */
- grpColIdx[keyno++] = te->resdom->resno;
+ grpColIdx[keyno++] = te->resno;
}
}
}
/*
- * make_groupsortplan
- * Add a Sort node to explicitly sort according to the GROUP BY clause.
+ * locate_grouping_columns
+ * Locate grouping columns in the tlist chosen by query_planner.
*
- * Note: the Sort node always just takes a copy of the subplan's tlist
- * plus ordering information. (This might seem inefficient if the
- * subplan contains complex GROUP BY expressions, but in fact Sort
- * does not evaluate its targetlist --- it only outputs the same
- * tuples in a new order. So the expressions we might be copying
- * are just dummies with no extra execution cost.)
- */
-static Plan *
-make_groupsortplan(Query *parse,
- List *groupClause,
- AttrNumber *grpColIdx,
- Plan *subplan)
-{
- List *sort_tlist = new_unsorted_tlist(subplan->targetlist);
- int keyno = 0;
- List *gl;
-
- foreach(gl, groupClause)
- {
- GroupClause *grpcl = (GroupClause *) lfirst(gl);
- TargetEntry *te = nth(grpColIdx[keyno] - 1, sort_tlist);
- Resdom *resdom = te->resdom;
-
- /*
- * Check for the possibility of duplicate group-by clauses ---
- * the parser should have removed 'em, but the Sort executor
- * will get terribly confused if any get through!
- */
- if (resdom->reskey == 0)
- {
- /* OK, insert the ordering info needed by the executor. */
- resdom->reskey = ++keyno;
- resdom->reskeyop = grpcl->sortop;
- }
- }
-
- Assert(keyno > 0);
-
- return (Plan *) make_sort(parse, sort_tlist, subplan, keyno);
-}
-
-/*
- * make_sortplan
- * Add a Sort node to implement an explicit ORDER BY clause.
+ * This is only needed if we don't use the sub_tlist chosen by
+ * make_subplanTargetList. We have to forget the column indexes found
+ * by that routine and re-locate the grouping vars in the real sub_tlist.
*/
-Plan *
-make_sortplan(Query *parse, List *tlist, Plan *plannode, List *sortcls)
+static void
+locate_grouping_columns(PlannerInfo *root,
+ List *tlist,
+ List *sub_tlist,
+ AttrNumber *groupColIdx)
{
- List *sort_tlist;
- List *i;
int keyno = 0;
+ ListCell *gl;
/*
- * First make a copy of the tlist so that we don't corrupt the
- * original.
+ * No work unless grouping.
*/
- sort_tlist = new_unsorted_tlist(tlist);
+ if (!root->parse->groupClause)
+ {
+ Assert(groupColIdx == NULL);
+ return;
+ }
+ Assert(groupColIdx != NULL);
- foreach(i, sortcls)
+ foreach(gl, root->parse->groupClause)
{
- SortClause *sortcl = (SortClause *) lfirst(i);
- TargetEntry *tle = get_sortgroupclause_tle(sortcl, sort_tlist);
- Resdom *resdom = tle->resdom;
+ GroupClause *grpcl = (GroupClause *) lfirst(gl);
+ Node *groupexpr = get_sortgroupclause_expr(grpcl, tlist);
+ TargetEntry *te = NULL;
+ ListCell *sl;
- /*
- * Check for the possibility of duplicate order-by clauses --- the
- * parser should have removed 'em, but the executor will get
- * terribly confused if any get through!
- */
- if (resdom->reskey == 0)
+ foreach(sl, sub_tlist)
{
- /* OK, insert the ordering info needed by the executor. */
- resdom->reskey = ++keyno;
- resdom->reskeyop = sortcl->sortop;
+ te = (TargetEntry *) lfirst(sl);
+ if (equal(groupexpr, te->expr))
+ break;
}
- }
-
- Assert(keyno > 0);
+ if (!sl)
+ elog(ERROR, "failed to locate grouping columns");
- return (Plan *) make_sort(parse, sort_tlist, plannode, keyno);
+ groupColIdx[keyno++] = te->resno;
+ }
}
/*
* We need to transpose sort key info from the orig_tlist into new_tlist.
* NOTE: this would not be good enough if we supported resjunk sort keys
* for results of set operations --- then, we'd need to project a whole
- * new tlist to evaluate the resjunk columns. For now, just elog if we
+ * new tlist to evaluate the resjunk columns. For now, just ereport if we
* find any resjunk columns in orig_tlist.
*/
static List *
postprocess_setop_tlist(List *new_tlist, List *orig_tlist)
{
- List *l;
+ ListCell *l;
+ ListCell *orig_tlist_item = list_head(orig_tlist);
foreach(l, new_tlist)
{
TargetEntry *orig_tle;
/* ignore resjunk columns in setop result */
- if (new_tle->resdom->resjunk)
+ if (new_tle->resjunk)
continue;
- Assert(orig_tlist != NIL);
- orig_tle = (TargetEntry *) lfirst(orig_tlist);
- orig_tlist = lnext(orig_tlist);
- if (orig_tle->resdom->resjunk)
- elog(ERROR, "postprocess_setop_tlist: resjunk output columns not implemented");
- Assert(new_tle->resdom->resno == orig_tle->resdom->resno);
- Assert(new_tle->resdom->restype == orig_tle->resdom->restype);
- new_tle->resdom->ressortgroupref = orig_tle->resdom->ressortgroupref;
+ Assert(orig_tlist_item != NULL);
+ orig_tle = (TargetEntry *) lfirst(orig_tlist_item);
+ orig_tlist_item = lnext(orig_tlist_item);
+ if (orig_tle->resjunk) /* should not happen */
+ elog(ERROR, "resjunk output columns are not implemented");
+ Assert(new_tle->resno == orig_tle->resno);
+ new_tle->ressortgroupref = orig_tle->ressortgroupref;
}
- if (orig_tlist != NIL)
- elog(ERROR, "postprocess_setop_tlist: resjunk output columns not implemented");
+ if (orig_tlist_item != NULL)
+ elog(ERROR, "resjunk output columns are not implemented");
return new_tlist;
}
projects / postgresql / blobdiff