* allpaths.c
* Routines to find possible search paths for processing a query
*
- * Portions Copyright (c) 1996-2005, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
- * $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.137 2005/10/15 02:49:19 momjian Exp $
+ * $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.175 2008/10/21 20:42:52 tgl Exp $
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
-#include "nodes/makefuncs.h"
+#include <math.h>
+
+#include "nodes/nodeFuncs.h"
#ifdef OPTIMIZER_DEBUG
#include "nodes/print.h"
#endif
#include "optimizer/paths.h"
#include "optimizer/plancat.h"
#include "optimizer/planner.h"
-#include "optimizer/predtest.h"
#include "optimizer/prep.h"
#include "optimizer/var.h"
-#include "parser/parsetree.h"
#include "parser/parse_clause.h"
-#include "parser/parse_expr.h"
+#include "parser/parsetree.h"
#include "rewrite/rewriteManip.h"
/* These parameters are set by GUC */
-bool constraint_exclusion = false;
bool enable_geqo = false; /* just in case GUC doesn't set it */
int geqo_threshold;
+/* Hook for plugins to replace standard_join_search() */
+join_search_hook_type join_search_hook = NULL;
+
static void set_base_rel_pathlists(PlannerInfo *root);
+static void set_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ Index rti, RangeTblEntry *rte);
static void set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
RangeTblEntry *rte);
-static void set_inherited_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
- Index rti, RangeTblEntry *rte,
- List *inheritlist);
+static void set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ Index rti, RangeTblEntry *rte);
+static void set_dummy_rel_pathlist(RelOptInfo *rel);
static void set_subquery_pathlist(PlannerInfo *root, RelOptInfo *rel,
Index rti, RangeTblEntry *rte);
static void set_function_pathlist(PlannerInfo *root, RelOptInfo *rel,
RangeTblEntry *rte);
-static RelOptInfo *make_one_rel_by_joins(PlannerInfo *root, int levels_needed,
- List *initial_rels);
+static void set_values_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ RangeTblEntry *rte);
+static void set_cte_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ RangeTblEntry *rte);
+static void set_worktable_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ RangeTblEntry *rte);
+static RelOptInfo *make_rel_from_joinlist(PlannerInfo *root, List *joinlist);
static bool subquery_is_pushdown_safe(Query *subquery, Query *topquery,
bool *differentTypes);
static bool recurse_pushdown_safe(Node *setOp, Query *topquery,
* single rel that represents the join of all base rels in the query.
*/
RelOptInfo *
-make_one_rel(PlannerInfo *root)
+make_one_rel(PlannerInfo *root, List *joinlist)
{
RelOptInfo *rel;
/*
* Generate access paths for the entire join tree.
*/
- Assert(root->parse->jointree != NULL &&
- IsA(root->parse->jointree, FromExpr));
-
- rel = make_fromexpr_rel(root, root->parse->jointree);
+ rel = make_rel_from_joinlist(root, joinlist);
/*
* The result should join all and only the query's base rels.
int num_base_rels = 0;
Index rti;
- for (rti = 1; rti < root->base_rel_array_size; rti++)
+ for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
- RelOptInfo *brel = root->base_rel_array[rti];
+ RelOptInfo *brel = root->simple_rel_array[rti];
if (brel == NULL)
continue;
{
Index rti;
- /*
- * Note: because we call expand_inherited_rtentry inside the loop, it's
- * quite possible for the base_rel_array to be enlarged while the loop
- * runs. Hence don't try to optimize the loop.
- */
- for (rti = 1; rti < root->base_rel_array_size; rti++)
+ for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
- RelOptInfo *rel = root->base_rel_array[rti];
- RangeTblEntry *rte;
- List *inheritlist;
+ RelOptInfo *rel = root->simple_rel_array[rti];
/* there may be empty slots corresponding to non-baserel RTEs */
if (rel == NULL)
if (rel->reloptkind != RELOPT_BASEREL)
continue;
- rte = rt_fetch(rti, root->parse->rtable);
+ set_rel_pathlist(root, rel, rti, root->simple_rte_array[rti]);
+ }
+}
- if (rel->rtekind == RTE_SUBQUERY)
- {
- /* Subquery --- generate a separate plan for it */
- set_subquery_pathlist(root, rel, rti, rte);
- }
- else if (rel->rtekind == RTE_FUNCTION)
- {
- /* RangeFunction --- generate a separate plan for it */
- set_function_pathlist(root, rel, rte);
- }
- else if ((inheritlist = expand_inherited_rtentry(root, rti)) != NIL)
- {
- /* Relation is root of an inheritance tree, process specially */
- set_inherited_rel_pathlist(root, rel, rti, rte, inheritlist);
- }
+/*
+ * set_rel_pathlist
+ * Build access paths for a base relation
+ */
+static void
+set_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ Index rti, RangeTblEntry *rte)
+{
+ if (rte->inh)
+ {
+ /* It's an "append relation", process accordingly */
+ set_append_rel_pathlist(root, rel, rti, rte);
+ }
+ else if (rel->rtekind == RTE_SUBQUERY)
+ {
+ /* Subquery --- generate a separate plan for it */
+ set_subquery_pathlist(root, rel, rti, rte);
+ }
+ else if (rel->rtekind == RTE_FUNCTION)
+ {
+ /* RangeFunction --- generate a suitable path for it */
+ set_function_pathlist(root, rel, rte);
+ }
+ else if (rel->rtekind == RTE_VALUES)
+ {
+ /* Values list --- generate a suitable path for it */
+ set_values_pathlist(root, rel, rte);
+ }
+ else if (rel->rtekind == RTE_CTE)
+ {
+ /* CTE reference --- generate a suitable path for it */
+ if (rte->self_reference)
+ set_worktable_pathlist(root, rel, rte);
else
- {
- /* Plain relation */
- set_plain_rel_pathlist(root, rel, rte);
- }
+ set_cte_pathlist(root, rel, rte);
+ }
+ else
+ {
+ /* Plain relation */
+ Assert(rel->rtekind == RTE_RELATION);
+ set_plain_rel_pathlist(root, rel, rte);
+ }
#ifdef OPTIMIZER_DEBUG
- debug_print_rel(root, rel);
+ debug_print_rel(root, rel);
#endif
- }
}
/*
static void
set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
{
+ /*
+ * If we can prove we don't need to scan the rel via constraint exclusion,
+ * set up a single dummy path for it. We only need to check for regular
+ * baserels; if it's an otherrel, CE was already checked in
+ * set_append_rel_pathlist().
+ */
+ if (rel->reloptkind == RELOPT_BASEREL &&
+ relation_excluded_by_constraints(root, rel, rte))
+ {
+ set_dummy_rel_pathlist(rel);
+ return;
+ }
+
/* Mark rel with estimated output rows, width, etc */
set_baserel_size_estimates(root, rel);
}
/*
- * set_inherited_rel_pathlist
- * Build access paths for a inheritance tree rooted at rel
- *
- * inheritlist is a list of RT indexes of all tables in the inheritance tree,
- * including a duplicate of the parent itself. Note we will not come here
- * unless there's at least one child in addition to the parent.
+ * set_append_rel_pathlist
+ * Build access paths for an "append relation"
*
- * NOTE: the passed-in rel and RTE will henceforth represent the appended
- * result of the whole inheritance tree. The members of inheritlist represent
- * the individual tables --- in particular, the inheritlist member that is a
- * duplicate of the parent RTE represents the parent table alone.
- * We will generate plans to scan the individual tables that refer to
- * the inheritlist RTEs, whereas Vars elsewhere in the plan tree that
- * refer to the original RTE are taken to refer to the append output.
- * In particular, this means we have separate RelOptInfos for the parent
- * table and for the append output, which is a good thing because they're
- * not the same size.
+ * The passed-in rel and RTE represent the entire append relation. The
+ * relation's contents are computed by appending together the output of
+ * the individual member relations. Note that in the inheritance case,
+ * the first member relation is actually the same table as is mentioned in
+ * the parent RTE ... but it has a different RTE and RelOptInfo. This is
+ * a good thing because their outputs are not the same size.
*/
static void
-set_inherited_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
- Index rti, RangeTblEntry *rte,
- List *inheritlist)
+set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ Index rti, RangeTblEntry *rte)
{
int parentRTindex = rti;
- Oid parentOID = rte->relid;
List *subpaths = NIL;
- ListCell *il;
+ double parent_rows;
+ double parent_size;
+ double *parent_attrsizes;
+ int nattrs;
+ ListCell *l;
/*
* XXX for now, can't handle inherited expansion of FOR UPDATE/SHARE; can
- * we do better?
+ * we do better? (This will take some redesign because the executor
+ * currently supposes that every rowMark relation is involved in every row
+ * returned by the query.)
*/
- if (list_member_int(root->parse->rowMarks, parentRTindex))
+ if (get_rowmark(root->parse, parentRTindex))
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("SELECT FOR UPDATE/SHARE is not supported for inheritance queries")));
/*
- * Initialize to compute size estimates for whole inheritance tree
+ * Initialize to compute size estimates for whole append relation.
+ *
+ * We handle width estimates by weighting the widths of different
+ * child rels proportionally to their number of rows. This is sensible
+ * because the use of width estimates is mainly to compute the total
+ * relation "footprint" if we have to sort or hash it. To do this,
+ * we sum the total equivalent size (in "double" arithmetic) and then
+ * divide by the total rowcount estimate. This is done separately for
+ * the total rel width and each attribute.
+ *
+ * Note: if you consider changing this logic, beware that child rels could
+ * have zero rows and/or width, if they were excluded by constraints.
*/
- rel->rows = 0;
- rel->width = 0;
+ parent_rows = 0;
+ parent_size = 0;
+ nattrs = rel->max_attr - rel->min_attr + 1;
+ parent_attrsizes = (double *) palloc0(nattrs * sizeof(double));
/*
- * Generate access paths for each table in the tree (parent AND children),
- * and pick the cheapest path for each table.
+ * Generate access paths for each member relation, and pick the cheapest
+ * path for each one.
*/
- foreach(il, inheritlist)
+ foreach(l, root->append_rel_list)
{
- int childRTindex = lfirst_int(il);
- RangeTblEntry *childrte;
- Oid childOID;
+ AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l);
+ int childRTindex;
+ RangeTblEntry *childRTE;
RelOptInfo *childrel;
+ Path *childpath;
ListCell *parentvars;
ListCell *childvars;
- childrte = rt_fetch(childRTindex, root->parse->rtable);
- childOID = childrte->relid;
+ /* append_rel_list contains all append rels; ignore others */
+ if (appinfo->parent_relid != parentRTindex)
+ continue;
+
+ childRTindex = appinfo->child_relid;
+ childRTE = root->simple_rte_array[childRTindex];
/*
- * Make a RelOptInfo for the child so we can do planning. Mark it as
- * an "other rel" since it will not be part of the main join tree.
+ * The child rel's RelOptInfo was already created during
+ * add_base_rels_to_query.
*/
- childrel = build_other_rel(root, childRTindex);
+ childrel = find_base_rel(root, childRTindex);
+ Assert(childrel->reloptkind == RELOPT_OTHER_MEMBER_REL);
/*
- * Copy the parent's targetlist and restriction quals to the child,
- * with attribute-number adjustment as needed. We don't bother to
- * copy the join quals, since we can't do any joining of the
- * individual tables. Also, we just zap attr_needed rather than
- * trying to adjust it; it won't be looked at in the child.
+ * We have to copy the parent's targetlist and quals to the child,
+ * with appropriate substitution of variables. However, only the
+ * baserestrictinfo quals are needed before we can check for
+ * constraint exclusion; so do that first and then check to see if we
+ * can disregard this child.
*/
- childrel->reltargetlist = (List *)
- adjust_inherited_attrs((Node *) rel->reltargetlist,
- parentRTindex,
- parentOID,
- childRTindex,
- childOID);
- childrel->attr_needed = NULL;
childrel->baserestrictinfo = (List *)
- adjust_inherited_attrs((Node *) rel->baserestrictinfo,
- parentRTindex,
- parentOID,
- childRTindex,
- childOID);
+ adjust_appendrel_attrs((Node *) rel->baserestrictinfo,
+ appinfo);
- /*
- * If we can prove we don't need to scan this child via constraint
- * exclusion, just ignore it. (We have to have converted the
- * baserestrictinfo Vars before we can make the test.)
- */
- if (constraint_exclusion)
+ if (relation_excluded_by_constraints(root, childrel, childRTE))
{
- List *constraint_pred;
-
- constraint_pred = get_relation_constraints(childOID, childrel);
-
/*
- * We do not currently enforce that CHECK constraints contain only
- * immutable functions, so it's necessary to check here. We
- * daren't draw conclusions from plan-time evaluation of
- * non-immutable functions.
+ * This child need not be scanned, so we can omit it from the
+ * appendrel. Mark it with a dummy cheapest-path though, in case
+ * best_appendrel_indexscan() looks at it later.
*/
- if (!contain_mutable_functions((Node *) constraint_pred))
- {
- /*
- * The constraints are effectively ANDed together, so we can
- * just try to refute the entire collection at once. This may
- * allow us to make proofs that would fail if we took them
- * individually.
- */
- if (predicate_refuted_by(constraint_pred,
- childrel->baserestrictinfo))
- continue;
- }
+ set_dummy_rel_pathlist(childrel);
+ continue;
}
+ /* CE failed, so finish copying targetlist and join quals */
+ childrel->joininfo = (List *)
+ adjust_appendrel_attrs((Node *) rel->joininfo,
+ appinfo);
+ childrel->reltargetlist = (List *)
+ adjust_appendrel_attrs((Node *) rel->reltargetlist,
+ appinfo);
+
/*
- * Compute the child's access paths, and save the cheapest.
+ * We have to make child entries in the EquivalenceClass data
+ * structures as well.
*/
- set_plain_rel_pathlist(root, childrel, childrte);
+ if (rel->has_eclass_joins)
+ {
+ add_child_rel_equivalences(root, appinfo, rel, childrel);
+ childrel->has_eclass_joins = true;
+ }
- subpaths = lappend(subpaths, childrel->cheapest_total_path);
+ /*
+ * Copy the parent's attr_needed data as well, with appropriate
+ * adjustment of relids and attribute numbers.
+ */
+ pfree(childrel->attr_needed);
+ childrel->attr_needed =
+ adjust_appendrel_attr_needed(rel, appinfo,
+ childrel->min_attr,
+ childrel->max_attr);
/*
- * Propagate size information from the child back to the parent. For
- * simplicity, we use the largest widths from any child as the parent
- * estimates.
+ * Compute the child's access paths, and add the cheapest one to the
+ * Append path we are constructing for the parent.
+ *
+ * It's possible that the child is itself an appendrel, in which case
+ * we can "cut out the middleman" and just add its child paths to our
+ * own list. (We don't try to do this earlier because we need to
+ * apply both levels of transformation to the quals.)
*/
- rel->rows += childrel->rows;
- if (childrel->width > rel->width)
- rel->width = childrel->width;
+ set_rel_pathlist(root, childrel, childRTindex, childRTE);
+
+ childpath = childrel->cheapest_total_path;
+ if (IsA(childpath, AppendPath))
+ subpaths = list_concat(subpaths,
+ ((AppendPath *) childpath)->subpaths);
+ else
+ subpaths = lappend(subpaths, childpath);
- forboth(parentvars, rel->reltargetlist,
- childvars, childrel->reltargetlist)
+ /*
+ * Accumulate size information from each child.
+ */
+ if (childrel->rows > 0)
{
- Var *parentvar = (Var *) lfirst(parentvars);
- Var *childvar = (Var *) lfirst(childvars);
+ parent_rows += childrel->rows;
+ parent_size += childrel->width * childrel->rows;
- if (IsA(parentvar, Var) &&IsA(childvar, Var))
+ forboth(parentvars, rel->reltargetlist,
+ childvars, childrel->reltargetlist)
{
- int pndx = parentvar->varattno - rel->min_attr;
- int cndx = childvar->varattno - childrel->min_attr;
+ Var *parentvar = (Var *) lfirst(parentvars);
+ Var *childvar = (Var *) lfirst(childvars);
- if (childrel->attr_widths[cndx] > rel->attr_widths[pndx])
- rel->attr_widths[pndx] = childrel->attr_widths[cndx];
+ /*
+ * Accumulate per-column estimates too. Whole-row Vars and
+ * PlaceHolderVars can be ignored here.
+ */
+ if (IsA(parentvar, Var) &&
+ IsA(childvar, Var))
+ {
+ int pndx = parentvar->varattno - rel->min_attr;
+ int cndx = childvar->varattno - childrel->min_attr;
+
+ parent_attrsizes[pndx] += childrel->attr_widths[cndx] * childrel->rows;
+ }
}
}
}
+ /*
+ * Save the finished size estimates.
+ */
+ rel->rows = parent_rows;
+ if (parent_rows > 0)
+ {
+ int i;
+
+ rel->width = rint(parent_size / parent_rows);
+ for (i = 0; i < nattrs; i++)
+ rel->attr_widths[i] = rint(parent_attrsizes[i] / parent_rows);
+ }
+ else
+ rel->width = 0; /* attr_widths should be zero already */
+
+ /*
+ * Set "raw tuples" count equal to "rows" for the appendrel; needed
+ * because some places assume rel->tuples is valid for any baserel.
+ */
+ rel->tuples = parent_rows;
+
+ pfree(parent_attrsizes);
+
/*
* Finally, build Append path and install it as the only access path for
* the parent rel. (Note: this is correct even if we have zero or one
set_cheapest(rel);
}
+/*
+ * set_dummy_rel_pathlist
+ * Build a dummy path for a relation that's been excluded by constraints
+ *
+ * Rather than inventing a special "dummy" path type, we represent this as an
+ * AppendPath with no members (see also IS_DUMMY_PATH macro).
+ */
+static void
+set_dummy_rel_pathlist(RelOptInfo *rel)
+{
+ /* Set dummy size estimates --- we leave attr_widths[] as zeroes */
+ rel->rows = 0;
+ rel->width = 0;
+
+ add_path(rel, (Path *) create_append_path(rel, NIL));
+
+ /* Select cheapest path (pretty easy in this case...) */
+ set_cheapest(rel);
+}
+
/* quick-and-dirty test to see if any joining is needed */
static bool
has_multiple_baserels(PlannerInfo *root)
int num_base_rels = 0;
Index rti;
- for (rti = 1; rti < root->base_rel_array_size; rti++)
+ for (rti = 1; rti < root->simple_rel_array_size; rti++)
{
- RelOptInfo *brel = root->base_rel_array[rti];
+ RelOptInfo *brel = root->simple_rel_array[rti];
if (brel == NULL)
continue;
Query *subquery = rte->subquery;
bool *differentTypes;
double tuple_fraction;
+ PlannerInfo *subroot;
List *pathkeys;
- List *subquery_pathkeys;
/* We need a workspace for keeping track of set-op type coercions */
differentTypes = (bool *)
* There are several cases where we cannot push down clauses. Restrictions
* involving the subquery are checked by subquery_is_pushdown_safe().
* Restrictions on individual clauses are checked by
- * qual_is_pushdown_safe().
+ * qual_is_pushdown_safe(). Also, we don't want to push down
+ * pseudoconstant clauses; better to have the gating node above the
+ * subquery.
*
- * Non-pushed-down clauses will get evaluated as qpquals of the SubqueryScan
- * node.
+ * Non-pushed-down clauses will get evaluated as qpquals of the
+ * SubqueryScan node.
*
* XXX Are there any cases where we want to make a policy decision not to
* push down a pushable qual, because it'd result in a worse plan?
RestrictInfo *rinfo = (RestrictInfo *) lfirst(l);
Node *clause = (Node *) rinfo->clause;
- if (qual_is_pushdown_safe(subquery, rti, clause, differentTypes))
+ if (!rinfo->pseudoconstant &&
+ qual_is_pushdown_safe(subquery, rti, clause, differentTypes))
{
/* Push it down */
subquery_push_qual(subquery, rte, rti, clause);
tuple_fraction = root->tuple_fraction;
/* Generate the plan for the subquery */
- rel->subplan = subquery_planner(subquery, tuple_fraction,
- &subquery_pathkeys);
+ rel->subplan = subquery_planner(root->glob, subquery,
+ root,
+ false, tuple_fraction,
+ &subroot);
+ rel->subrtable = subroot->parse->rtable;
/* Copy number of output rows from subplan */
rel->tuples = rel->subplan->plan_rows;
set_baserel_size_estimates(root, rel);
/* Convert subquery pathkeys to outer representation */
- pathkeys = convert_subquery_pathkeys(root, rel, subquery_pathkeys);
+ pathkeys = convert_subquery_pathkeys(root, rel, subroot->query_pathkeys);
/* Generate appropriate path */
add_path(rel, create_subqueryscan_path(rel, pathkeys));
}
/*
- * make_fromexpr_rel
- * Build access paths for a FromExpr jointree node.
+ * set_values_pathlist
+ * Build the (single) access path for a VALUES RTE
*/
-RelOptInfo *
-make_fromexpr_rel(PlannerInfo *root, FromExpr *from)
+static void
+set_values_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
+{
+ /* Mark rel with estimated output rows, width, etc */
+ set_values_size_estimates(root, rel);
+
+ /* Generate appropriate path */
+ add_path(rel, create_valuesscan_path(root, rel));
+
+ /* Select cheapest path (pretty easy in this case...) */
+ set_cheapest(rel);
+}
+
+/*
+ * set_cte_pathlist
+ * Build the (single) access path for a non-self-reference CTE RTE
+ */
+static void
+set_cte_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
+{
+ Plan *cteplan;
+ PlannerInfo *cteroot;
+ Index levelsup;
+ int ndx;
+ ListCell *lc;
+ int plan_id;
+
+ /*
+ * Find the referenced CTE, and locate the plan previously made for it.
+ */
+ levelsup = rte->ctelevelsup;
+ cteroot = root;
+ while (levelsup-- > 0)
+ {
+ cteroot = cteroot->parent_root;
+ if (!cteroot) /* shouldn't happen */
+ elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename);
+ }
+ /*
+ * Note: cte_plan_ids can be shorter than cteList, if we are still working
+ * on planning the CTEs (ie, this is a side-reference from another CTE).
+ * So we mustn't use forboth here.
+ */
+ ndx = 0;
+ foreach(lc, cteroot->parse->cteList)
+ {
+ CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
+
+ if (strcmp(cte->ctename, rte->ctename) == 0)
+ break;
+ ndx++;
+ }
+ if (lc == NULL) /* shouldn't happen */
+ elog(ERROR, "could not find CTE \"%s\"", rte->ctename);
+ if (ndx >= list_length(cteroot->cte_plan_ids))
+ elog(ERROR, "could not find plan for CTE \"%s\"", rte->ctename);
+ plan_id = list_nth_int(cteroot->cte_plan_ids, ndx);
+ Assert(plan_id > 0);
+ cteplan = (Plan *) list_nth(root->glob->subplans, plan_id - 1);
+
+ /* Mark rel with estimated output rows, width, etc */
+ set_cte_size_estimates(root, rel, cteplan);
+
+ /* Generate appropriate path */
+ add_path(rel, create_ctescan_path(root, rel));
+
+ /* Select cheapest path (pretty easy in this case...) */
+ set_cheapest(rel);
+}
+
+/*
+ * set_worktable_pathlist
+ * Build the (single) access path for a self-reference CTE RTE
+ */
+static void
+set_worktable_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
+{
+ Plan *cteplan;
+ PlannerInfo *cteroot;
+ Index levelsup;
+
+ /*
+ * We need to find the non-recursive term's plan, which is in the plan
+ * level that's processing the recursive UNION, which is one level
+ * *below* where the CTE comes from.
+ */
+ levelsup = rte->ctelevelsup;
+ if (levelsup == 0) /* shouldn't happen */
+ elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename);
+ levelsup--;
+ cteroot = root;
+ while (levelsup-- > 0)
+ {
+ cteroot = cteroot->parent_root;
+ if (!cteroot) /* shouldn't happen */
+ elog(ERROR, "bad levelsup for CTE \"%s\"", rte->ctename);
+ }
+ cteplan = cteroot->non_recursive_plan;
+ if (!cteplan) /* shouldn't happen */
+ elog(ERROR, "could not find plan for CTE \"%s\"", rte->ctename);
+
+ /* Mark rel with estimated output rows, width, etc */
+ set_cte_size_estimates(root, rel, cteplan);
+
+ /* Generate appropriate path */
+ add_path(rel, create_worktablescan_path(root, rel));
+
+ /* Select cheapest path (pretty easy in this case...) */
+ set_cheapest(rel);
+}
+
+/*
+ * make_rel_from_joinlist
+ * Build access paths using a "joinlist" to guide the join path search.
+ *
+ * See comments for deconstruct_jointree() for definition of the joinlist
+ * data structure.
+ */
+static RelOptInfo *
+make_rel_from_joinlist(PlannerInfo *root, List *joinlist)
{
int levels_needed;
- List *initial_rels = NIL;
- ListCell *jt;
+ List *initial_rels;
+ ListCell *jl;
/*
- * Count the number of child jointree nodes. This is the depth of the
+ * Count the number of child joinlist nodes. This is the depth of the
* dynamic-programming algorithm we must employ to consider all ways of
* joining the child nodes.
*/
- levels_needed = list_length(from->fromlist);
+ levels_needed = list_length(joinlist);
if (levels_needed <= 0)
return NULL; /* nothing to do? */
/*
- * Construct a list of rels corresponding to the child jointree nodes.
+ * Construct a list of rels corresponding to the child joinlist nodes.
* This may contain both base rels and rels constructed according to
- * explicit JOIN directives.
+ * sub-joinlists.
*/
- foreach(jt, from->fromlist)
+ initial_rels = NIL;
+ foreach(jl, joinlist)
{
- Node *jtnode = (Node *) lfirst(jt);
+ Node *jlnode = (Node *) lfirst(jl);
+ RelOptInfo *thisrel;
+
+ if (IsA(jlnode, RangeTblRef))
+ {
+ int varno = ((RangeTblRef *) jlnode)->rtindex;
- initial_rels = lappend(initial_rels,
- make_jointree_rel(root, jtnode));
+ thisrel = find_base_rel(root, varno);
+ }
+ else if (IsA(jlnode, List))
+ {
+ /* Recurse to handle subproblem */
+ thisrel = make_rel_from_joinlist(root, (List *) jlnode);
+ }
+ else
+ {
+ elog(ERROR, "unrecognized joinlist node type: %d",
+ (int) nodeTag(jlnode));
+ thisrel = NULL; /* keep compiler quiet */
+ }
+
+ initial_rels = lappend(initial_rels, thisrel);
}
if (levels_needed == 1)
{
/*
- * Single jointree node, so we're done.
+ * Single joinlist node, so we're done.
*/
return (RelOptInfo *) linitial(initial_rels);
}
{
/*
* Consider the different orders in which we could join the rels,
- * using either GEQO or regular optimizer.
+ * using a plugin, GEQO, or the regular join search code.
+ *
+ * We put the initial_rels list into a PlannerInfo field because
+ * has_legal_joinclause() needs to look at it (ugly :-().
*/
- if (enable_geqo && levels_needed >= geqo_threshold)
+ root->initial_rels = initial_rels;
+
+ if (join_search_hook)
+ return (*join_search_hook) (root, levels_needed, initial_rels);
+ else if (enable_geqo && levels_needed >= geqo_threshold)
return geqo(root, levels_needed, initial_rels);
else
- return make_one_rel_by_joins(root, levels_needed, initial_rels);
+ return standard_join_search(root, levels_needed, initial_rels);
}
}
/*
- * make_one_rel_by_joins
- * Find all possible joinpaths for a query by successively finding ways
+ * standard_join_search
+ * Find possible joinpaths for a query by successively finding ways
* to join component relations into join relations.
*
* 'levels_needed' is the number of iterations needed, ie, the number of
*
* 'initial_rels' is a list of RelOptInfo nodes for each independent
* jointree item. These are the components to be joined together.
+ * Note that levels_needed == list_length(initial_rels).
*
* Returns the final level of join relations, i.e., the relation that is
* the result of joining all the original relations together.
+ * At least one implementation path must be provided for this relation and
+ * all required sub-relations.
+ *
+ * To support loadable plugins that modify planner behavior by changing the
+ * join searching algorithm, we provide a hook variable that lets a plugin
+ * replace or supplement this function. Any such hook must return the same
+ * final join relation as the standard code would, but it might have a
+ * different set of implementation paths attached, and only the sub-joinrels
+ * needed for these paths need have been instantiated.
+ *
+ * Note to plugin authors: the functions invoked during standard_join_search()
+ * modify root->join_rel_list and root->join_rel_hash. If you want to do more
+ * than one join-order search, you'll probably need to save and restore the
+ * original states of those data structures. See geqo_eval() for an example.
*/
-static RelOptInfo *
-make_one_rel_by_joins(PlannerInfo *root, int levels_needed, List *initial_rels)
+RelOptInfo *
+standard_join_search(PlannerInfo *root, int levels_needed, List *initial_rels)
{
List **joinitems;
int lev;
* level, and build paths for making each one from every available
* pair of lower-level relations.
*/
- joinitems[lev] = make_rels_by_joins(root, lev, joinitems);
+ joinitems[lev] = join_search_one_level(root, lev, joinitems);
/*
* Do cleanup work on each just-processed rel.
* Compare tlist's datatypes against the list of set-operation result types.
* For any items that are different, mark the appropriate element of
* differentTypes[] to show that this column will have type conversions.
+ *
+ * We don't have to care about typmods here: the only allowed difference
+ * between set-op input and output typmods is input is a specific typmod
+ * and output is -1, and that does not require a coercion.
*/
static void
compare_tlist_datatypes(List *tlist, List *colTypes,
* it will work correctly: sublinks will already have been transformed into
* subplans in the qual, but not in the subquery).
*
- * 2. The qual must not refer to any subquery output columns that were
+ * 2. The qual must not refer to the whole-row output of the subquery
+ * (since there is no easy way to name that within the subquery itself).
+ *
+ * 3. The qual must not refer to any subquery output columns that were
* found to have inconsistent types across a set operation tree by
* subquery_is_pushdown_safe().
*
- * 3. If the subquery uses DISTINCT ON, we must not push down any quals that
+ * 4. If the subquery uses DISTINCT ON, we must not push down any quals that
* refer to non-DISTINCT output columns, because that could change the set
- * of rows returned. This condition is vacuous for DISTINCT, because then
- * there are no non-DISTINCT output columns, but unfortunately it's fairly
- * expensive to tell the difference between DISTINCT and DISTINCT ON in the
- * parsetree representation. It's cheaper to just make sure all the Vars
- * in the qual refer to DISTINCT columns.
+ * of rows returned. (This condition is vacuous for DISTINCT, because then
+ * there are no non-DISTINCT output columns, so we needn't check. But note
+ * we are assuming that the qual can't distinguish values that the DISTINCT
+ * operator sees as equal. This is a bit shaky but we have no way to test
+ * for the case, and it's unlikely enough that we shouldn't refuse the
+ * optimization just because it could theoretically happen.)
*
- * 4. We must not push down any quals that refer to subselect outputs that
+ * 5. We must not push down any quals that refer to subselect outputs that
* return sets, else we'd introduce functions-returning-sets into the
* subquery's WHERE/HAVING quals.
+ *
+ * 6. We must not push down any quals that refer to subselect outputs that
+ * contain volatile functions, for fear of introducing strange results due
+ * to multiple evaluation of a volatile function.
*/
static bool
qual_is_pushdown_safe(Query *subquery, Index rti, Node *qual,
* Examine all Vars used in clause; since it's a restriction clause, all
* such Vars must refer to subselect output columns.
*/
- vars = pull_var_clause(qual, false);
+ vars = pull_var_clause(qual, true);
foreach(vl, vars)
{
Var *var = (Var *) lfirst(vl);
TargetEntry *tle;
+ /*
+ * XXX Punt if we find any PlaceHolderVars in the restriction clause.
+ * It's not clear whether a PHV could safely be pushed down, and even
+ * less clear whether such a situation could arise in any cases of
+ * practical interest anyway. So for the moment, just refuse to push
+ * down.
+ */
+ if (!IsA(var, Var))
+ {
+ safe = false;
+ break;
+ }
+
Assert(var->varno == rti);
+ /* Check point 2 */
+ if (var->varattno == 0)
+ {
+ safe = false;
+ break;
+ }
+
/*
* We use a bitmapset to avoid testing the same attno more than once.
* (NB: this only works because subquery outputs can't have negative
continue;
tested = bms_add_member(tested, var->varattno);
- /* Check point 2 */
+ /* Check point 3 */
if (differentTypes[var->varattno])
{
safe = false;
Assert(tle != NULL);
Assert(!tle->resjunk);
- /* If subquery uses DISTINCT or DISTINCT ON, check point 3 */
- if (subquery->distinctClause != NIL &&
- !targetIsInSortList(tle, subquery->distinctClause))
+ /* If subquery uses DISTINCT ON, check point 4 */
+ if (subquery->hasDistinctOn &&
+ !targetIsInSortList(tle, InvalidOid, subquery->distinctClause))
{
/* non-DISTINCT column, so fail */
safe = false;
break;
}
- /* Refuse functions returning sets (point 4) */
+ /* Refuse functions returning sets (point 5) */
if (expression_returns_set((Node *) tle->expr))
{
safe = false;
break;
}
+
+ /* Refuse volatile functions (point 6) */
+ if (contain_volatile_functions((Node *) tle->expr))
+ {
+ safe = false;
+ break;
+ }
}
list_free(vars);
break;
case T_ResultPath:
ptype = "Result";
- subpath = ((ResultPath *) path)->subpath;
break;
case T_MaterialPath:
ptype = "Material";
projects / postgresql / blobdiff