* core.c
* Routines copied from PostgreSQL core distribution.
*
+ * The main purpose of this files is having access to static functions in core.
+ * Another purpose is tweaking functions behavior by replacing part of them by
+ * macro definitions. See at the end of pg_hint_plan.c for details. Anyway,
+ * this file *must* contain required functions without making any change.
+ *
+ * This file contains the following functions from corresponding files.
+ *
* src/backend/optimizer/path/allpaths.c
- * standard_join_search()
- * set_plain_rel_pathlist()
- * set_append_rel_pathlist()
- * accumulate_append_subpath()
- * set_dummy_rel_pathlist()
+ *
+ * static functions:
+ * set_plain_rel_pathlist()
+ * add_paths_to_append_rel()
+ * try_partitionwise_join()
+ *
+ * public functions:
+ * standard_join_search(): This funcion is not static. The reason for
+ * including this function is make_rels_by_clause_joins. In order to
+ * avoid generating apparently unwanted join combination, we decided to
+ * change the behavior of make_join_rel, which is called under this
+ * function.
*
* src/backend/optimizer/path/joinrels.c
- * join_search_one_level()
+ *
+ * public functions:
+ * join_search_one_level(): We have to modify this to call my definition of
+ * make_rels_by_clause_joins.
+ *
+ * static functions:
* make_rels_by_clause_joins()
* make_rels_by_clauseless_joins()
* join_is_legal()
* has_join_restriction()
- * is_dummy_rel()
- * mark_dummy_rel()
* restriction_is_constant_false()
+ * update_child_rel_info()
+ * build_child_join_sjinfo()
+ * get_matching_part_pairs()
+ * compute_partition_bounds()
*
- * Portions Copyright (c) 1996-2011, PostgreSQL Global Development Group
+ *
+ * Portions Copyright (c) 1996-2020, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*-------------------------------------------------------------------------
*/
+static void populate_joinrel_with_paths(PlannerInfo *root, RelOptInfo *rel1,
+ RelOptInfo *rel2, RelOptInfo *joinrel,
+ SpecialJoinInfo *sjinfo, List *restrictlist);
+
+/*
+ * set_plain_rel_pathlist
+ * Build access paths for a plain relation (no subquery, no inheritance)
+ */
+static void
+set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
+{
+ Relids required_outer;
+
+ /*
+ * We don't support pushing join clauses into the quals of a seqscan, but
+ * it could still have required parameterization due to LATERAL refs in
+ * its tlist.
+ */
+ required_outer = rel->lateral_relids;
+
+ /* Consider sequential scan */
+ add_path(rel, create_seqscan_path(root, rel, required_outer, 0));
+
+ /* If appropriate, consider parallel sequential scan */
+ if (rel->consider_parallel && required_outer == NULL)
+ create_plain_partial_paths(root, rel);
+
+ /* Consider index scans */
+ create_index_paths(root, rel);
+
+ /* Consider TID scans */
+ create_tidscan_paths(root, rel);
+}
+
+
+/*
+ * set_append_rel_pathlist
+ * Build access paths for an "append relation"
+ */
+static void
+set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
+ Index rti, RangeTblEntry *rte)
+{
+ int parentRTindex = rti;
+ List *live_childrels = NIL;
+ ListCell *l;
+
+ /*
+ * Generate access paths for each member relation, and remember the
+ * non-dummy children.
+ */
+ foreach(l, root->append_rel_list)
+ {
+ AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l);
+ int childRTindex;
+ RangeTblEntry *childRTE;
+ RelOptInfo *childrel;
+
+ /* append_rel_list contains all append rels; ignore others */
+ if (appinfo->parent_relid != parentRTindex)
+ continue;
+
+ /* Re-locate the child RTE and RelOptInfo */
+ childRTindex = appinfo->child_relid;
+ childRTE = root->simple_rte_array[childRTindex];
+ childrel = root->simple_rel_array[childRTindex];
+
+ /*
+ * If set_append_rel_size() decided the parent appendrel was
+ * parallel-unsafe at some point after visiting this child rel, we
+ * need to propagate the unsafety marking down to the child, so that
+ * we don't generate useless partial paths for it.
+ */
+ if (!rel->consider_parallel)
+ childrel->consider_parallel = false;
+
+ /*
+ * Compute the child's access paths.
+ */
+ set_rel_pathlist(root, childrel, childRTindex, childRTE);
+
+ /*
+ * If child is dummy, ignore it.
+ */
+ if (IS_DUMMY_REL(childrel))
+ continue;
+
+ /* Bubble up childrel's partitioned children. */
+ if (rel->part_scheme)
+ rel->partitioned_child_rels =
+ list_concat(rel->partitioned_child_rels,
+ childrel->partitioned_child_rels);
+
+ /*
+ * Child is live, so add it to the live_childrels list for use below.
+ */
+ live_childrels = lappend(live_childrels, childrel);
+ }
+
+ /* Add paths to the append relation. */
+ add_paths_to_append_rel(root, rel, live_childrels);
+}
+
+
/*
* standard_join_search
* Find possible joinpaths for a query by successively finding ways
* independent jointree items in the query. This is > 1.
*
* 'initial_rels' is a list of RelOptInfo nodes for each independent
- * jointree item. These are the components to be joined together.
+ * 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
* 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
+ * 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.
*/
join_search_one_level(root, lev);
/*
- * Do cleanup work on each just-processed rel.
+ * Run generate_partitionwise_join_paths() and generate_gather_paths()
+ * for each just-processed joinrel. We could not do this earlier
+ * because both regular and partial paths can get added to a
+ * particular joinrel at multiple times within join_search_one_level.
+ *
+ * After that, we're done creating paths for the joinrel, so run
+ * set_cheapest().
*/
foreach(lc, root->join_rel_level[lev])
{
rel = (RelOptInfo *) lfirst(lc);
+ /* Create paths for partitionwise joins. */
+ generate_partitionwise_join_paths(root, rel);
+
+ /*
+ * Except for the topmost scan/join rel, consider gathering
+ * partial paths. We'll do the same for the topmost scan/join rel
+ * once we know the final targetlist (see grouping_planner).
+ */
+ if (lev < levels_needed)
+ generate_useful_gather_paths(root, rel, false);
+
/* Find and save the cheapest paths for this rel */
set_cheapest(rel);
}
/*
- * set_plain_rel_pathlist
- * Build access paths for a plain relation (no subquery, no inheritance)
- */
-static void
-set_plain_rel_pathlist(PlannerInfo *root, RelOptInfo *rel, RangeTblEntry *rte)
-{
- /* Consider sequential scan */
- add_path(rel, create_seqscan_path(root, rel));
-
- /* Consider index scans */
- create_index_paths(root, rel);
-
- /* Consider TID scans */
- create_tidscan_paths(root, rel);
-
- /* Now find the cheapest of the paths for this rel */
- set_cheapest(rel);
-}
-
-/*
- * set_append_rel_pathlist
- * Build access paths for an "append relation"
- *
- * 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.
+ * create_plain_partial_paths
+ * Build partial access paths for parallel scan of a plain relation
*/
static void
-set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
- Index rti, RangeTblEntry *rte)
+create_plain_partial_paths(PlannerInfo *root, RelOptInfo *rel)
{
- int parentRTindex = rti;
- List *live_childrels = NIL;
- List *subpaths = NIL;
- List *all_child_pathkeys = NIL;
- double parent_rows;
- double parent_size;
- double *parent_attrsizes;
- int nattrs;
- ListCell *l;
-
- /*
- * 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.
- */
- parent_rows = 0;
- parent_size = 0;
- nattrs = rel->max_attr - rel->min_attr + 1;
- parent_attrsizes = (double *) palloc0(nattrs * sizeof(double));
+ int parallel_workers;
- /*
- * Generate access paths for each member relation, and pick the cheapest
- * path for each one.
- */
- foreach(l, root->append_rel_list)
- {
- AppendRelInfo *appinfo = (AppendRelInfo *) lfirst(l);
- int childRTindex;
- RangeTblEntry *childRTE;
- RelOptInfo *childrel;
- List *childquals;
- Node *childqual;
- ListCell *lcp;
- ListCell *parentvars;
- ListCell *childvars;
+ parallel_workers = compute_parallel_worker(rel, rel->pages, -1,
+ max_parallel_workers_per_gather);
- /* 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];
-
- /*
- * The child rel's RelOptInfo was already created during
- * add_base_rels_to_query.
- */
- childrel = find_base_rel(root, childRTindex);
- Assert(childrel->reloptkind == RELOPT_OTHER_MEMBER_REL);
-
- /*
- * 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.
- *
- * As of 8.4, the child rel's targetlist might contain non-Var
- * expressions, which means that substitution into the quals could
- * produce opportunities for const-simplification, and perhaps even
- * pseudoconstant quals. To deal with this, we strip the RestrictInfo
- * nodes, do the substitution, do const-simplification, and then
- * reconstitute the RestrictInfo layer.
- */
- childquals = get_all_actual_clauses(rel->baserestrictinfo);
- childquals = (List *) adjust_appendrel_attrs((Node *) childquals,
- appinfo);
- childqual = eval_const_expressions(root, (Node *)
- make_ands_explicit(childquals));
- if (childqual && IsA(childqual, Const) &&
- (((Const *) childqual)->constisnull ||
- !DatumGetBool(((Const *) childqual)->constvalue)))
- {
- /*
- * Restriction reduces to constant FALSE or constant NULL after
- * substitution, so this child need not be scanned.
- */
- set_dummy_rel_pathlist(childrel);
- continue;
- }
- childquals = make_ands_implicit((Expr *) childqual);
- childquals = make_restrictinfos_from_actual_clauses(root,
- childquals);
- childrel->baserestrictinfo = childquals;
-
- if (relation_excluded_by_constraints(root, childrel, childRTE))
- {
- /*
- * 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.
- */
- set_dummy_rel_pathlist(childrel);
- continue;
- }
-
- /*
- * CE failed, so finish copying/modifying targetlist and join quals.
- *
- * Note: the resulting childrel->reltargetlist may contain arbitrary
- * expressions, which normally would not occur in a reltargetlist.
- * That is okay because nothing outside of this routine will look at
- * the child rel's reltargetlist. We do have to cope with the case
- * while constructing attr_widths estimates below, though.
- */
- childrel->joininfo = (List *)
- adjust_appendrel_attrs((Node *) rel->joininfo,
- appinfo);
- childrel->reltargetlist = (List *)
- adjust_appendrel_attrs((Node *) rel->reltargetlist,
- appinfo);
-
- /*
- * We have to make child entries in the EquivalenceClass data
- * structures as well. This is needed either if the parent
- * participates in some eclass joins (because we will want to consider
- * inner-indexscan joins on the individual children) or if the parent
- * has useful pathkeys (because we should try to build MergeAppend
- * paths that produce those sort orderings).
- */
- if (rel->has_eclass_joins || has_useful_pathkeys(root, rel))
- add_child_rel_equivalences(root, appinfo, rel, childrel);
- childrel->has_eclass_joins = rel->has_eclass_joins;
-
- /*
- * Note: we could compute appropriate attr_needed data for the child's
- * variables, by transforming the parent's attr_needed through the
- * translated_vars mapping. However, currently there's no need
- * because attr_needed is only examined for base relations not
- * otherrels. So we just leave the child's attr_needed empty.
- */
-
- /* Remember which childrels are live, for MergeAppend logic below */
- live_childrels = lappend(live_childrels, childrel);
-
- /*
- * Compute the child's access paths, and add the cheapest one to the
- * Append path we are constructing for the parent.
- */
- set_rel_pathlist(root, childrel, childRTindex, childRTE);
-
- subpaths = accumulate_append_subpath(subpaths,
- childrel->cheapest_total_path);
-
- /*
- * Collect a list of all the available path orderings for all the
- * children. We use this as a heuristic to indicate which sort
- * orderings we should build MergeAppend paths for.
- */
- foreach(lcp, childrel->pathlist)
- {
- Path *childpath = (Path *) lfirst(lcp);
- List *childkeys = childpath->pathkeys;
- ListCell *lpk;
- bool found = false;
-
- /* Ignore unsorted paths */
- if (childkeys == NIL)
- continue;
-
- /* Have we already seen this ordering? */
- foreach(lpk, all_child_pathkeys)
- {
- List *existing_pathkeys = (List *) lfirst(lpk);
-
- if (compare_pathkeys(existing_pathkeys,
- childkeys) == PATHKEYS_EQUAL)
- {
- found = true;
- break;
- }
- }
- if (!found)
- {
- /* No, so add it to all_child_pathkeys */
- all_child_pathkeys = lappend(all_child_pathkeys, childkeys);
- }
- }
-
- /*
- * Accumulate size information from each child.
- */
- if (childrel->rows > 0)
- {
- parent_rows += childrel->rows;
- parent_size += childrel->width * childrel->rows;
-
- /*
- * Accumulate per-column estimates too. We need not do anything
- * for PlaceHolderVars in the parent list. If child expression
- * isn't a Var, or we didn't record a width estimate for it, we
- * have to fall back on a datatype-based estimate.
- *
- * By construction, child's reltargetlist is 1-to-1 with parent's.
- */
- forboth(parentvars, rel->reltargetlist,
- childvars, childrel->reltargetlist)
- {
- Var *parentvar = (Var *) lfirst(parentvars);
- Node *childvar = (Node *) lfirst(childvars);
-
- if (IsA(parentvar, Var))
- {
- int pndx = parentvar->varattno - rel->min_attr;
- int32 child_width = 0;
-
- if (IsA(childvar, Var))
- {
- int cndx = ((Var *) childvar)->varattno - childrel->min_attr;
-
- child_width = childrel->attr_widths[cndx];
- }
- if (child_width <= 0)
- child_width = get_typavgwidth(exprType(childvar),
- exprTypmod(childvar));
- Assert(child_width > 0);
- parent_attrsizes[pndx] += child_width * 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);
-
- /*
- * Next, build an unordered Append path for the rel. (Note: this is
- * correct even if we have zero or one live subpath due to constraint
- * exclusion.)
- */
- add_path(rel, (Path *) create_append_path(rel, subpaths));
-
- /*
- * Next, build MergeAppend paths based on the collected list of child
- * pathkeys. We consider both cheapest-startup and cheapest-total cases,
- * ie, for each interesting ordering, collect all the cheapest startup
- * subpaths and all the cheapest total paths, and build a MergeAppend path
- * for each list.
- */
- foreach(l, all_child_pathkeys)
- {
- List *pathkeys = (List *) lfirst(l);
- List *startup_subpaths = NIL;
- List *total_subpaths = NIL;
- bool startup_neq_total = false;
- ListCell *lcr;
-
- /* Select the child paths for this ordering... */
- foreach(lcr, live_childrels)
- {
- RelOptInfo *childrel = (RelOptInfo *) lfirst(lcr);
- Path *cheapest_startup,
- *cheapest_total;
-
- /* Locate the right paths, if they are available. */
- cheapest_startup =
- get_cheapest_path_for_pathkeys(childrel->pathlist,
- pathkeys,
- STARTUP_COST);
- cheapest_total =
- get_cheapest_path_for_pathkeys(childrel->pathlist,
- pathkeys,
- TOTAL_COST);
-
- /*
- * If we can't find any paths with the right order just add the
- * cheapest-total path; we'll have to sort it.
- */
- if (cheapest_startup == NULL)
- cheapest_startup = childrel->cheapest_total_path;
- if (cheapest_total == NULL)
- cheapest_total = childrel->cheapest_total_path;
-
- /*
- * Notice whether we actually have different paths for the
- * "cheapest" and "total" cases; frequently there will be no point
- * in two create_merge_append_path() calls.
- */
- if (cheapest_startup != cheapest_total)
- startup_neq_total = true;
-
- startup_subpaths =
- accumulate_append_subpath(startup_subpaths, cheapest_startup);
- total_subpaths =
- accumulate_append_subpath(total_subpaths, cheapest_total);
- }
-
- /* ... and build the MergeAppend paths */
- add_path(rel, (Path *) create_merge_append_path(root,
- rel,
- startup_subpaths,
- pathkeys));
- if (startup_neq_total)
- add_path(rel, (Path *) create_merge_append_path(root,
- rel,
- total_subpaths,
- pathkeys));
- }
-
- /* Select cheapest path */
- set_cheapest(rel);
-}
-
-/*
- * accumulate_append_subpath
- * Add a subpath to the list being built for an Append or MergeAppend
- *
- * It's possible that the child is itself an Append path, 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.)
- */
-static List *
-accumulate_append_subpath(List *subpaths, Path *path)
-{
- if (IsA(path, AppendPath))
- {
- AppendPath *apath = (AppendPath *) path;
+ /* If any limit was set to zero, the user doesn't want a parallel scan. */
+ if (parallel_workers <= 0)
+ return;
- /* list_copy is important here to avoid sharing list substructure */
- return list_concat(subpaths, list_copy(apath->subpaths));
- }
- else
- return lappend(subpaths, path);
+ /* Add an unordered partial path based on a parallel sequential scan. */
+ add_partial_path(rel, create_seqscan_path(root, rel, NULL, parallel_workers));
}
-/*
- * 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);
-}
/*
* join_search_one_level
* We prefer to join using join clauses, but if we find a rel of level-1
* members that has no join clauses, we will generate Cartesian-product
* joins against all initial rels not already contained in it.
- *
- * In the first pass (level == 2), we try to join each initial rel to each
- * initial rel that appears later in joinrels[1]. (The mirror-image joins
- * are handled automatically by make_join_rel.) In later passes, we try
- * to join rels of size level-1 from joinrels[level-1] to each initial rel
- * in joinrels[1].
*/
foreach(r, joinrels[level - 1])
{
RelOptInfo *old_rel = (RelOptInfo *) lfirst(r);
- ListCell *other_rels;
-
- if (level == 2)
- other_rels = lnext(r); /* only consider remaining initial
- * rels */
- else
- other_rels = list_head(joinrels[1]); /* consider all initial
- * rels */
if (old_rel->joininfo != NIL || old_rel->has_eclass_joins ||
has_join_restriction(root, old_rel))
{
/*
- * Note that if all available join clauses for this rel require
- * more than one other rel, we will fail to make any joins against
- * it here. In most cases that's OK; it'll be considered by
- * "bushy plan" join code in a higher-level pass where we have
- * those other rels collected into a join rel.
+ * There are join clauses or join order restrictions relevant to
+ * this rel, so consider joins between this rel and (only) those
+ * initial rels it is linked to by a clause or restriction.
*
- * See also the last-ditch case below.
+ * At level 2 this condition is symmetric, so there is no need to
+ * look at initial rels before this one in the list; we already
+ * considered such joins when we were at the earlier rel. (The
+ * mirror-image joins are handled automatically by make_join_rel.)
+ * In later passes (level > 2), we join rels of the previous level
+ * to each initial rel they don't already include but have a join
+ * clause or restriction with.
*/
+ List *other_rels_list;
+ ListCell *other_rels;
+
+ if (level == 2) /* consider remaining initial rels */
+ {
+ other_rels_list = joinrels[level - 1];
+ other_rels = lnext(other_rels_list, r);
+ }
+ else /* consider all initial rels */
+ {
+ other_rels_list = joinrels[1];
+ other_rels = list_head(other_rels_list);
+ }
+
make_rels_by_clause_joins(root,
old_rel,
+ other_rels_list,
other_rels);
}
else
* Oops, we have a relation that is not joined to any other
* relation, either directly or by join-order restrictions.
* Cartesian product time.
+ *
+ * We consider a cartesian product with each not-already-included
+ * initial rel, whether it has other join clauses or not. At
+ * level 2, if there are two or more clauseless initial rels, we
+ * will redundantly consider joining them in both directions; but
+ * such cases aren't common enough to justify adding complexity to
+ * avoid the duplicated effort.
*/
make_rels_by_clauseless_joins(root,
old_rel,
- other_rels);
+ joinrels[1]);
}
}
foreach(r, joinrels[k])
{
RelOptInfo *old_rel = (RelOptInfo *) lfirst(r);
+ List *other_rels_list;
ListCell *other_rels;
ListCell *r2;
/*
- * We can ignore clauseless joins here, *except* when they
+ * We can ignore relations without join clauses here, unless they
* participate in join-order restrictions --- then we might have
* to force a bushy join plan.
*/
continue;
if (k == other_level)
- other_rels = lnext(r); /* only consider remaining rels */
+ {
+ /* only consider remaining rels */
+ other_rels_list = joinrels[k];
+ other_rels = lnext(other_rels_list, r);
+ }
else
- other_rels = list_head(joinrels[other_level]);
+ {
+ other_rels_list = joinrels[other_level];
+ other_rels = list_head(other_rels_list);
+ }
- for_each_cell(r2, other_rels)
+ for_each_cell(r2, other_rels_list, other_rels)
{
RelOptInfo *new_rel = (RelOptInfo *) lfirst(r2);
{
/*
* OK, we can build a rel of the right level from this
- * pair of rels. Do so if there is at least one usable
- * join clause or a relevant join restriction.
+ * pair of rels. Do so if there is at least one relevant
+ * join clause or join order restriction.
*/
if (have_relevant_joinclause(root, old_rel, new_rel) ||
have_join_order_restriction(root, old_rel, new_rel))
}
}
- /*
+ /*----------
* Last-ditch effort: if we failed to find any usable joins so far, force
* a set of cartesian-product joins to be generated. This handles the
* special case where all the available rels have join clauses but we
- * cannot use any of those clauses yet. An example is
+ * cannot use any of those clauses yet. This can only happen when we are
+ * considering a join sub-problem (a sub-joinlist) and all the rels in the
+ * sub-problem have only join clauses with rels outside the sub-problem.
+ * An example is
*
- * SELECT * FROM a,b,c WHERE (a.f1 + b.f2 + c.f3) = 0;
+ * SELECT ... FROM a INNER JOIN b ON TRUE, c, d, ...
+ * WHERE a.w = c.x and b.y = d.z;
*
- * The join clause will be usable at level 3, but at level 2 we have no
- * choice but to make cartesian joins. We consider only left-sided and
- * right-sided cartesian joins in this case (no bushy).
+ * If the "a INNER JOIN b" sub-problem does not get flattened into the
+ * upper level, we must be willing to make a cartesian join of a and b;
+ * but the code above will not have done so, because it thought that both
+ * a and b have joinclauses. We consider only left-sided and right-sided
+ * cartesian joins in this case (no bushy).
+ *----------
*/
if (joinrels[level] == NIL)
{
foreach(r, joinrels[level - 1])
{
RelOptInfo *old_rel = (RelOptInfo *) lfirst(r);
- ListCell *other_rels;
-
- if (level == 2)
- other_rels = lnext(r); /* only consider remaining initial
- * rels */
- else
- other_rels = list_head(joinrels[1]); /* consider all initial
- * rels */
make_rels_by_clauseless_joins(root,
old_rel,
- other_rels);
+ joinrels[1]);
}
/*----------
* When special joins are involved, there may be no legal way
- * to make an N-way join for some values of N. For example consider
+ * to make an N-way join for some values of N. For example consider
*
* SELECT ... FROM t1 WHERE
* x IN (SELECT ... FROM t2,t3 WHERE ...) AND
* to accept failure at level 4 and go on to discover a workable
* bushy plan at level 5.
*
- * However, if there are no special joins then join_is_legal() should
- * never fail, and so the following sanity check is useful.
+ * However, if there are no special joins and no lateral references
+ * then join_is_legal() should never fail, and so the following sanity
+ * check is useful.
*----------
*/
- if (joinrels[level] == NIL && root->join_info_list == NIL)
+ if (joinrels[level] == NIL &&
+ root->join_info_list == NIL &&
+ !root->hasLateralRTEs)
elog(ERROR, "failed to build any %d-way joins", level);
}
}
+
/*
* make_rels_by_clause_joins
* Build joins between the given relation 'old_rel' and other relations
* automatically ensures that each new joinrel is only added to the list once.
*
* 'old_rel' is the relation entry for the relation to be joined
- * 'other_rels': the first cell in a linked list containing the other
+ * 'other_rels_list': a list containing the other
* rels to be considered for joining
+ * 'other_rels': the first cell to be considered
*
* Currently, this is only used with initial rels in other_rels, but it
* will work for joining to joinrels too.
static void
make_rels_by_clause_joins(PlannerInfo *root,
RelOptInfo *old_rel,
+ List *other_rels_list,
ListCell *other_rels)
{
ListCell *l;
- for_each_cell(l, other_rels)
+ for_each_cell(l, other_rels_list, other_rels)
{
RelOptInfo *other_rel = (RelOptInfo *) lfirst(l);
}
}
+
/*
* make_rels_by_clauseless_joins
* Given a relation 'old_rel' and a list of other relations
* The join rels are returned in root->join_rel_level[join_cur_level].
*
* 'old_rel' is the relation entry for the relation to be joined
- * 'other_rels': the first cell of a linked list containing the
- * other rels to be considered for joining
+ * 'other_rels': a list containing the other rels to be considered for joining
*
* Currently, this is only used with initial rels in other_rels, but it would
* work for joining to joinrels too.
static void
make_rels_by_clauseless_joins(PlannerInfo *root,
RelOptInfo *old_rel,
- ListCell *other_rels)
+ List *other_rels)
{
ListCell *l;
- for_each_cell(l, other_rels)
+ foreach(l, other_rels)
{
RelOptInfo *other_rel = (RelOptInfo *) lfirst(l);
}
}
+
/*
* join_is_legal
* Determine whether a proposed join is legal given the query's
*
* On success, *sjinfo_p is set to NULL if this is to be a plain inner join,
* else it's set to point to the associated SpecialJoinInfo node. Also,
- * *reversed_p is set TRUE if the given relations need to be swapped to
+ * *reversed_p is set true if the given relations need to be swapped to
* match the SpecialJoinInfo node.
*/
static bool
SpecialJoinInfo *match_sjinfo;
bool reversed;
bool unique_ified;
- bool is_valid_inner;
+ bool must_be_leftjoin;
ListCell *l;
/*
- * Ensure output params are set on failure return. This is just to
+ * Ensure output params are set on failure return. This is just to
* suppress uninitialized-variable warnings from overly anal compilers.
*/
*sjinfo_p = NULL;
/*
* If we have any special joins, the proposed join might be illegal; and
- * in any case we have to determine its join type. Scan the join info
- * list for conflicts.
+ * in any case we have to determine its join type. Scan the join info
+ * list for matches and conflicts.
*/
match_sjinfo = NULL;
reversed = false;
unique_ified = false;
- is_valid_inner = true;
+ must_be_leftjoin = false;
foreach(l, root->join_info_list)
{
* If one input contains min_lefthand and the other contains
* min_righthand, then we can perform the SJ at this join.
*
- * Barf if we get matches to more than one SJ (is that possible?)
+ * Reject if we get matches to more than one SJ; that implies we're
+ * considering something that's not really valid.
*/
if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) &&
bms_is_subset(sjinfo->min_righthand, rel2->relids))
}
else
{
- /*----------
- * Otherwise, the proposed join overlaps the RHS but isn't
- * a valid implementation of this SJ. It might still be
- * a legal join, however. If both inputs overlap the RHS,
- * assume that it's OK. Since the inputs presumably got past
- * this function's checks previously, they can't overlap the
- * LHS and their violations of the RHS boundary must represent
- * SJs that have been determined to commute with this one.
- * We have to allow this to work correctly in cases like
- * (a LEFT JOIN (b JOIN (c LEFT JOIN d)))
- * when the c/d join has been determined to commute with the join
- * to a, and hence d is not part of min_righthand for the upper
- * join. It should be legal to join b to c/d but this will appear
- * as a violation of the upper join's RHS.
- * Furthermore, if one input overlaps the RHS and the other does
- * not, we should still allow the join if it is a valid
- * implementation of some other SJ. We have to allow this to
- * support the associative identity
- * (a LJ b on Pab) LJ c ON Pbc = a LJ (b LJ c ON Pbc) on Pab
- * since joining B directly to C violates the lower SJ's RHS.
- * We assume that make_outerjoininfo() set things up correctly
- * so that we'll only match to some SJ if the join is valid.
- * Set flag here to check at bottom of loop.
- *----------
+ /*
+ * Otherwise, the proposed join overlaps the RHS but isn't a valid
+ * implementation of this SJ. But don't panic quite yet: the RHS
+ * violation might have occurred previously, in one or both input
+ * relations, in which case we must have previously decided that
+ * it was OK to commute some other SJ with this one. If we need
+ * to perform this join to finish building up the RHS, rejecting
+ * it could lead to not finding any plan at all. (This can occur
+ * because of the heuristics elsewhere in this file that postpone
+ * clauseless joins: we might not consider doing a clauseless join
+ * within the RHS until after we've performed other, validly
+ * commutable SJs with one or both sides of the clauseless join.)
+ * This consideration boils down to the rule that if both inputs
+ * overlap the RHS, we can allow the join --- they are either
+ * fully within the RHS, or represent previously-allowed joins to
+ * rels outside it.
*/
- if (sjinfo->jointype != JOIN_SEMI &&
- bms_overlap(rel1->relids, sjinfo->min_righthand) &&
+ if (bms_overlap(rel1->relids, sjinfo->min_righthand) &&
bms_overlap(rel2->relids, sjinfo->min_righthand))
- {
- /* seems OK */
- Assert(!bms_overlap(joinrelids, sjinfo->min_lefthand));
- }
- else
- is_valid_inner = false;
+ continue; /* assume valid previous violation of RHS */
+
+ /*
+ * The proposed join could still be legal, but only if we're
+ * allowed to associate it into the RHS of this SJ. That means
+ * this SJ must be a LEFT join (not SEMI or ANTI, and certainly
+ * not FULL) and the proposed join must not overlap the LHS.
+ */
+ if (sjinfo->jointype != JOIN_LEFT ||
+ bms_overlap(joinrelids, sjinfo->min_lefthand))
+ return false; /* invalid join path */
+
+ /*
+ * To be valid, the proposed join must be a LEFT join; otherwise
+ * it can't associate into this SJ's RHS. But we may not yet have
+ * found the SpecialJoinInfo matching the proposed join, so we
+ * can't test that yet. Remember the requirement for later.
+ */
+ must_be_leftjoin = true;
}
}
/*
- * Fail if violated some SJ's RHS and didn't match to another SJ. However,
- * "matching" to a semijoin we are implementing by unique-ification
- * doesn't count (think: it's really an inner join).
+ * Fail if violated any SJ's RHS and didn't match to a LEFT SJ: the
+ * proposed join can't associate into an SJ's RHS.
+ *
+ * Also, fail if the proposed join's predicate isn't strict; we're
+ * essentially checking to see if we can apply outer-join identity 3, and
+ * that's a requirement. (This check may be redundant with checks in
+ * make_outerjoininfo, but I'm not quite sure, and it's cheap to test.)
*/
- if (!is_valid_inner &&
- (match_sjinfo == NULL || unique_ified))
+ if (must_be_leftjoin &&
+ (match_sjinfo == NULL ||
+ match_sjinfo->jointype != JOIN_LEFT ||
+ !match_sjinfo->lhs_strict))
return false; /* invalid join path */
+ /*
+ * We also have to check for constraints imposed by LATERAL references.
+ */
+ if (root->hasLateralRTEs)
+ {
+ bool lateral_fwd;
+ bool lateral_rev;
+ Relids join_lateral_rels;
+
+ /*
+ * The proposed rels could each contain lateral references to the
+ * other, in which case the join is impossible. If there are lateral
+ * references in just one direction, then the join has to be done with
+ * a nestloop with the lateral referencer on the inside. If the join
+ * matches an SJ that cannot be implemented by such a nestloop, the
+ * join is impossible.
+ *
+ * Also, if the lateral reference is only indirect, we should reject
+ * the join; whatever rel(s) the reference chain goes through must be
+ * joined to first.
+ *
+ * Another case that might keep us from building a valid plan is the
+ * implementation restriction described by have_dangerous_phv().
+ */
+ lateral_fwd = bms_overlap(rel1->relids, rel2->lateral_relids);
+ lateral_rev = bms_overlap(rel2->relids, rel1->lateral_relids);
+ if (lateral_fwd && lateral_rev)
+ return false; /* have lateral refs in both directions */
+ if (lateral_fwd)
+ {
+ /* has to be implemented as nestloop with rel1 on left */
+ if (match_sjinfo &&
+ (reversed ||
+ unique_ified ||
+ match_sjinfo->jointype == JOIN_FULL))
+ return false; /* not implementable as nestloop */
+ /* check there is a direct reference from rel2 to rel1 */
+ if (!bms_overlap(rel1->relids, rel2->direct_lateral_relids))
+ return false; /* only indirect refs, so reject */
+ /* check we won't have a dangerous PHV */
+ if (have_dangerous_phv(root, rel1->relids, rel2->lateral_relids))
+ return false; /* might be unable to handle required PHV */
+ }
+ else if (lateral_rev)
+ {
+ /* has to be implemented as nestloop with rel2 on left */
+ if (match_sjinfo &&
+ (!reversed ||
+ unique_ified ||
+ match_sjinfo->jointype == JOIN_FULL))
+ return false; /* not implementable as nestloop */
+ /* check there is a direct reference from rel1 to rel2 */
+ if (!bms_overlap(rel2->relids, rel1->direct_lateral_relids))
+ return false; /* only indirect refs, so reject */
+ /* check we won't have a dangerous PHV */
+ if (have_dangerous_phv(root, rel2->relids, rel1->lateral_relids))
+ return false; /* might be unable to handle required PHV */
+ }
+
+ /*
+ * LATERAL references could also cause problems later on if we accept
+ * this join: if the join's minimum parameterization includes any rels
+ * that would have to be on the inside of an outer join with this join
+ * rel, then it's never going to be possible to build the complete
+ * query using this join. We should reject this join not only because
+ * it'll save work, but because if we don't, the clauseless-join
+ * heuristics might think that legality of this join means that some
+ * other join rel need not be formed, and that could lead to failure
+ * to find any plan at all. We have to consider not only rels that
+ * are directly on the inner side of an OJ with the joinrel, but also
+ * ones that are indirectly so, so search to find all such rels.
+ */
+ join_lateral_rels = min_join_parameterization(root, joinrelids,
+ rel1, rel2);
+ if (join_lateral_rels)
+ {
+ Relids join_plus_rhs = bms_copy(joinrelids);
+ bool more;
+
+ do
+ {
+ more = false;
+ foreach(l, root->join_info_list)
+ {
+ SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l);
+
+ /* ignore full joins --- their ordering is predetermined */
+ if (sjinfo->jointype == JOIN_FULL)
+ continue;
+
+ if (bms_overlap(sjinfo->min_lefthand, join_plus_rhs) &&
+ !bms_is_subset(sjinfo->min_righthand, join_plus_rhs))
+ {
+ join_plus_rhs = bms_add_members(join_plus_rhs,
+ sjinfo->min_righthand);
+ more = true;
+ }
+ }
+ } while (more);
+ if (bms_overlap(join_plus_rhs, join_lateral_rels))
+ return false; /* will not be able to join to some RHS rel */
+ }
+ }
+
/* Otherwise, it's a valid join */
*sjinfo_p = match_sjinfo;
*reversed_p = reversed;
return true;
}
+
/*
* has_join_restriction
- * Detect whether the specified relation has join-order restrictions
- * due to being inside an outer join or an IN (sub-SELECT).
+ * Detect whether the specified relation has join-order restrictions,
+ * due to being inside an outer join or an IN (sub-SELECT),
+ * or participating in any LATERAL references or multi-rel PHVs.
*
* Essentially, this tests whether have_join_order_restriction() could
* succeed with this rel and some other one. It's OK if we sometimes
- * say "true" incorrectly. (Therefore, we don't bother with the relatively
+ * say "true" incorrectly. (Therefore, we don't bother with the relatively
* expensive has_legal_joinclause test.)
*/
static bool
{
ListCell *l;
+ if (rel->lateral_relids != NULL || rel->lateral_referencers != NULL)
+ return true;
+
+ foreach(l, root->placeholder_list)
+ {
+ PlaceHolderInfo *phinfo = (PlaceHolderInfo *) lfirst(l);
+
+ if (bms_is_subset(rel->relids, phinfo->ph_eval_at) &&
+ !bms_equal(rel->relids, phinfo->ph_eval_at))
+ return true;
+ }
+
foreach(l, root->join_info_list)
{
SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l);
}
/*
- * is_dummy_rel --- has relation been proven empty?
- *
- * If so, it will have a single path that is dummy.
- */
-static bool
-is_dummy_rel(RelOptInfo *rel)
-{
- return (rel->cheapest_total_path != NULL &&
- IS_DUMMY_PATH(rel->cheapest_total_path));
-}
-
-/*
- * Mark a relation as proven empty.
- *
- * During GEQO planning, this can get invoked more than once on the same
- * baserel struct, so it's worth checking to see if the rel is already marked
- * dummy.
- *
- * Also, when called during GEQO join planning, we are in a short-lived
- * memory context. We must make sure that the dummy path attached to a
- * baserel survives the GEQO cycle, else the baserel is trashed for future
- * GEQO cycles. On the other hand, when we are marking a joinrel during GEQO,
- * we don't want the dummy path to clutter the main planning context. Upshot
- * is that the best solution is to explicitly make the dummy path in the same
- * context the given RelOptInfo is in.
- */
-static void
-mark_dummy_rel(RelOptInfo *rel)
-{
- MemoryContext oldcontext;
-
- /* Already marked? */
- if (is_dummy_rel(rel))
- return;
-
- /* No, so choose correct context to make the dummy path in */
- oldcontext = MemoryContextSwitchTo(GetMemoryChunkContext(rel));
-
- /* Set dummy size estimate */
- rel->rows = 0;
-
- /* Evict any previously chosen paths */
- rel->pathlist = NIL;
-
- /* Set up the dummy path */
- add_path(rel, (Path *) create_append_path(rel, NIL));
-
- /* Set or update cheapest_total_path */
- set_cheapest(rel);
-
- MemoryContextSwitchTo(oldcontext);
-}
-
-/*
* restriction_is_constant_false --- is a restrictlist just FALSE?
*
* In cases where a qual is provably constant FALSE, eval_const_expressions
* decide there's no match for an outer row, which is pretty stupid. So,
* we need to detect the case.
*
- * If only_pushed_down is TRUE, then consider only pushed-down quals.
+ * If only_pushed_down is true, then consider only quals that are pushed-down
+ * from the point of view of the joinrel.
*/
static bool
-restriction_is_constant_false(List *restrictlist, bool only_pushed_down)
+restriction_is_constant_false(List *restrictlist,
+ RelOptInfo *joinrel,
+ bool only_pushed_down)
{
ListCell *lc;
*/
foreach(lc, restrictlist)
{
- RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc);
+ RestrictInfo *rinfo = lfirst_node(RestrictInfo, lc);
- Assert(IsA(rinfo, RestrictInfo));
- if (only_pushed_down && !rinfo->is_pushed_down)
+ if (only_pushed_down && !RINFO_IS_PUSHED_DOWN(rinfo, joinrel->relids))
continue;
if (rinfo->clause && IsA(rinfo->clause, Const))
}
return false;
}
+
+/*
+ * Construct the SpecialJoinInfo for a child-join by translating
+ * SpecialJoinInfo for the join between parents. left_relids and right_relids
+ * are the relids of left and right side of the join respectively.
+ */
+static SpecialJoinInfo *
+build_child_join_sjinfo(PlannerInfo *root, SpecialJoinInfo *parent_sjinfo,
+ Relids left_relids, Relids right_relids)
+{
+ SpecialJoinInfo *sjinfo = makeNode(SpecialJoinInfo);
+ AppendRelInfo **left_appinfos;
+ int left_nappinfos;
+ AppendRelInfo **right_appinfos;
+ int right_nappinfos;
+
+ memcpy(sjinfo, parent_sjinfo, sizeof(SpecialJoinInfo));
+ left_appinfos = find_appinfos_by_relids(root, left_relids,
+ &left_nappinfos);
+ right_appinfos = find_appinfos_by_relids(root, right_relids,
+ &right_nappinfos);
+
+ sjinfo->min_lefthand = adjust_child_relids(sjinfo->min_lefthand,
+ left_nappinfos, left_appinfos);
+ sjinfo->min_righthand = adjust_child_relids(sjinfo->min_righthand,
+ right_nappinfos,
+ right_appinfos);
+ sjinfo->syn_lefthand = adjust_child_relids(sjinfo->syn_lefthand,
+ left_nappinfos, left_appinfos);
+ sjinfo->syn_righthand = adjust_child_relids(sjinfo->syn_righthand,
+ right_nappinfos,
+ right_appinfos);
+ sjinfo->semi_rhs_exprs = (List *) adjust_appendrel_attrs(root,
+ (Node *) sjinfo->semi_rhs_exprs,
+ right_nappinfos,
+ right_appinfos);
+
+ pfree(left_appinfos);
+ pfree(right_appinfos);
+
+ return sjinfo;
+}
+
+/*
+ * get_matching_part_pairs
+ * Generate pairs of partitions to be joined from inputs
+ */
+static void
+get_matching_part_pairs(PlannerInfo *root, RelOptInfo *joinrel,
+ RelOptInfo *rel1, RelOptInfo *rel2,
+ List **parts1, List **parts2)
+{
+ bool rel1_is_simple = IS_SIMPLE_REL(rel1);
+ bool rel2_is_simple = IS_SIMPLE_REL(rel2);
+ int cnt_parts;
+
+ *parts1 = NIL;
+ *parts2 = NIL;
+
+ for (cnt_parts = 0; cnt_parts < joinrel->nparts; cnt_parts++)
+ {
+ RelOptInfo *child_joinrel = joinrel->part_rels[cnt_parts];
+ RelOptInfo *child_rel1;
+ RelOptInfo *child_rel2;
+ Relids child_relids1;
+ Relids child_relids2;
+
+ /*
+ * If this segment of the join is empty, it means that this segment
+ * was ignored when previously creating child-join paths for it in
+ * try_partitionwise_join() as it would not contribute to the join
+ * result, due to one or both inputs being empty; add NULL to each of
+ * the given lists so that this segment will be ignored again in that
+ * function.
+ */
+ if (!child_joinrel)
+ {
+ *parts1 = lappend(*parts1, NULL);
+ *parts2 = lappend(*parts2, NULL);
+ continue;
+ }
+
+ /*
+ * Get a relids set of partition(s) involved in this join segment that
+ * are from the rel1 side.
+ */
+ child_relids1 = bms_intersect(child_joinrel->relids,
+ rel1->all_partrels);
+ Assert(bms_num_members(child_relids1) == bms_num_members(rel1->relids));
+
+ /*
+ * Get a child rel for rel1 with the relids. Note that we should have
+ * the child rel even if rel1 is a join rel, because in that case the
+ * partitions specified in the relids would have matching/overlapping
+ * boundaries, so the specified partitions should be considered as
+ * ones to be joined when planning partitionwise joins of rel1,
+ * meaning that the child rel would have been built by the time we get
+ * here.
+ */
+ if (rel1_is_simple)
+ {
+ int varno = bms_singleton_member(child_relids1);
+
+ child_rel1 = find_base_rel(root, varno);
+ }
+ else
+ child_rel1 = find_join_rel(root, child_relids1);
+ Assert(child_rel1);
+
+ /*
+ * Get a relids set of partition(s) involved in this join segment that
+ * are from the rel2 side.
+ */
+ child_relids2 = bms_intersect(child_joinrel->relids,
+ rel2->all_partrels);
+ Assert(bms_num_members(child_relids2) == bms_num_members(rel2->relids));
+
+ /*
+ * Get a child rel for rel2 with the relids. See above comments.
+ */
+ if (rel2_is_simple)
+ {
+ int varno = bms_singleton_member(child_relids2);
+
+ child_rel2 = find_base_rel(root, varno);
+ }
+ else
+ child_rel2 = find_join_rel(root, child_relids2);
+ Assert(child_rel2);
+
+ /*
+ * The join of rel1 and rel2 is legal, so is the join of the child
+ * rels obtained above; add them to the given lists as a join pair
+ * producing this join segment.
+ */
+ *parts1 = lappend(*parts1, child_rel1);
+ *parts2 = lappend(*parts2, child_rel2);
+ }
+}
+
+
+/*
+ * compute_partition_bounds
+ * Compute the partition bounds for a join rel from those for inputs
+ */
+static void
+compute_partition_bounds(PlannerInfo *root, RelOptInfo *rel1,
+ RelOptInfo *rel2, RelOptInfo *joinrel,
+ SpecialJoinInfo *parent_sjinfo,
+ List **parts1, List **parts2)
+{
+ /*
+ * If we don't have the partition bounds for the join rel yet, try to
+ * compute those along with pairs of partitions to be joined.
+ */
+ if (joinrel->nparts == -1)
+ {
+ PartitionScheme part_scheme = joinrel->part_scheme;
+ PartitionBoundInfo boundinfo = NULL;
+ int nparts = 0;
+
+ Assert(joinrel->boundinfo == NULL);
+ Assert(joinrel->part_rels == NULL);
+
+ /*
+ * See if the partition bounds for inputs are exactly the same, in
+ * which case we don't need to work hard: the join rel have the same
+ * partition bounds as inputs, and the partitions with the same
+ * cardinal positions form the pairs.
+ *
+ * Note: even in cases where one or both inputs have merged bounds, it
+ * would be possible for both the bounds to be exactly the same, but
+ * it seems unlikely to be worth the cycles to check.
+ */
+ if (!rel1->partbounds_merged &&
+ !rel2->partbounds_merged &&
+ rel1->nparts == rel2->nparts &&
+ partition_bounds_equal(part_scheme->partnatts,
+ part_scheme->parttyplen,
+ part_scheme->parttypbyval,
+ rel1->boundinfo, rel2->boundinfo))
+ {
+ boundinfo = rel1->boundinfo;
+ nparts = rel1->nparts;
+ }
+ else
+ {
+ /* Try merging the partition bounds for inputs. */
+ boundinfo = partition_bounds_merge(part_scheme->partnatts,
+ part_scheme->partsupfunc,
+ part_scheme->partcollation,
+ rel1, rel2,
+ parent_sjinfo->jointype,
+ parts1, parts2);
+ if (boundinfo == NULL)
+ {
+ joinrel->nparts = 0;
+ return;
+ }
+ nparts = list_length(*parts1);
+ joinrel->partbounds_merged = true;
+ }
+
+ Assert(nparts > 0);
+ joinrel->boundinfo = boundinfo;
+ joinrel->nparts = nparts;
+ joinrel->part_rels =
+ (RelOptInfo **) palloc0(sizeof(RelOptInfo *) * nparts);
+ }
+ else
+ {
+ Assert(joinrel->nparts > 0);
+ Assert(joinrel->boundinfo);
+ Assert(joinrel->part_rels);
+
+ /*
+ * If the join rel's partbounds_merged flag is true, it means inputs
+ * are not guaranteed to have the same partition bounds, therefore we
+ * can't assume that the partitions at the same cardinal positions
+ * form the pairs; let get_matching_part_pairs() generate the pairs.
+ * Otherwise, nothing to do since we can assume that.
+ */
+ if (joinrel->partbounds_merged)
+ {
+ get_matching_part_pairs(root, joinrel, rel1, rel2,
+ parts1, parts2);
+ Assert(list_length(*parts1) == joinrel->nparts);
+ Assert(list_length(*parts2) == joinrel->nparts);
+ }
+ }
+}
+
+
+/*
+ * Assess whether join between given two partitioned relations can be broken
+ * down into joins between matching partitions; a technique called
+ * "partitionwise join"
+ *
+ * Partitionwise join is possible when a. Joining relations have same
+ * partitioning scheme b. There exists an equi-join between the partition keys
+ * of the two relations.
+ *
+ * Partitionwise join is planned as follows (details: optimizer/README.)
+ *
+ * 1. Create the RelOptInfos for joins between matching partitions i.e
+ * child-joins and add paths to them.
+ *
+ * 2. Construct Append or MergeAppend paths across the set of child joins.
+ * This second phase is implemented by generate_partitionwise_join_paths().
+ *
+ * The RelOptInfo, SpecialJoinInfo and restrictlist for each child join are
+ * obtained by translating the respective parent join structures.
+ */
+static void
+try_partitionwise_join(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2,
+ RelOptInfo *joinrel, SpecialJoinInfo *parent_sjinfo,
+ List *parent_restrictlist)
+{
+ bool rel1_is_simple = IS_SIMPLE_REL(rel1);
+ bool rel2_is_simple = IS_SIMPLE_REL(rel2);
+ List *parts1 = NIL;
+ List *parts2 = NIL;
+ ListCell *lcr1 = NULL;
+ ListCell *lcr2 = NULL;
+ int cnt_parts;
+
+ /* Guard against stack overflow due to overly deep partition hierarchy. */
+ check_stack_depth();
+
+ /* Nothing to do, if the join relation is not partitioned. */
+ if (joinrel->part_scheme == NULL || joinrel->nparts == 0)
+ return;
+
+ /* The join relation should have consider_partitionwise_join set. */
+ Assert(joinrel->consider_partitionwise_join);
+
+ /*
+ * We can not perform partitionwise join if either of the joining
+ * relations is not partitioned.
+ */
+ if (!IS_PARTITIONED_REL(rel1) || !IS_PARTITIONED_REL(rel2))
+ return;
+
+ Assert(REL_HAS_ALL_PART_PROPS(rel1) && REL_HAS_ALL_PART_PROPS(rel2));
+
+ /* The joining relations should have consider_partitionwise_join set. */
+ Assert(rel1->consider_partitionwise_join &&
+ rel2->consider_partitionwise_join);
+
+ /*
+ * The partition scheme of the join relation should match that of the
+ * joining relations.
+ */
+ Assert(joinrel->part_scheme == rel1->part_scheme &&
+ joinrel->part_scheme == rel2->part_scheme);
+
+ Assert(!(joinrel->partbounds_merged && (joinrel->nparts <= 0)));
+
+ compute_partition_bounds(root, rel1, rel2, joinrel, parent_sjinfo,
+ &parts1, &parts2);
+
+ if (joinrel->partbounds_merged)
+ {
+ lcr1 = list_head(parts1);
+ lcr2 = list_head(parts2);
+ }
+
+ /*
+ * Create child-join relations for this partitioned join, if those don't
+ * exist. Add paths to child-joins for a pair of child relations
+ * corresponding to the given pair of parent relations.
+ */
+ for (cnt_parts = 0; cnt_parts < joinrel->nparts; cnt_parts++)
+ {
+ RelOptInfo *child_rel1;
+ RelOptInfo *child_rel2;
+ bool rel1_empty;
+ bool rel2_empty;
+ SpecialJoinInfo *child_sjinfo;
+ List *child_restrictlist;
+ RelOptInfo *child_joinrel;
+ Relids child_joinrelids;
+ AppendRelInfo **appinfos;
+ int nappinfos;
+
+ if (joinrel->partbounds_merged)
+ {
+ child_rel1 = lfirst_node(RelOptInfo, lcr1);
+ child_rel2 = lfirst_node(RelOptInfo, lcr2);
+ lcr1 = lnext(parts1, lcr1);
+ lcr2 = lnext(parts2, lcr2);
+ }
+ else
+ {
+ child_rel1 = rel1->part_rels[cnt_parts];
+ child_rel2 = rel2->part_rels[cnt_parts];
+ }
+
+ rel1_empty = (child_rel1 == NULL || IS_DUMMY_REL(child_rel1));
+ rel2_empty = (child_rel2 == NULL || IS_DUMMY_REL(child_rel2));
+
+ /*
+ * Check for cases where we can prove that this segment of the join
+ * returns no rows, due to one or both inputs being empty (including
+ * inputs that have been pruned away entirely). If so just ignore it.
+ * These rules are equivalent to populate_joinrel_with_paths's rules
+ * for dummy input relations.
+ */
+ switch (parent_sjinfo->jointype)
+ {
+ case JOIN_INNER:
+ case JOIN_SEMI:
+ if (rel1_empty || rel2_empty)
+ continue; /* ignore this join segment */
+ break;
+ case JOIN_LEFT:
+ case JOIN_ANTI:
+ if (rel1_empty)
+ continue; /* ignore this join segment */
+ break;
+ case JOIN_FULL:
+ if (rel1_empty && rel2_empty)
+ continue; /* ignore this join segment */
+ break;
+ default:
+ /* other values not expected here */
+ elog(ERROR, "unrecognized join type: %d",
+ (int) parent_sjinfo->jointype);
+ break;
+ }
+
+ /*
+ * If a child has been pruned entirely then we can't generate paths
+ * for it, so we have to reject partitionwise joining unless we were
+ * able to eliminate this partition above.
+ */
+ if (child_rel1 == NULL || child_rel2 == NULL)
+ {
+ /*
+ * Mark the joinrel as unpartitioned so that later functions treat
+ * it correctly.
+ */
+ joinrel->nparts = 0;
+ return;
+ }
+
+ /*
+ * If a leaf relation has consider_partitionwise_join=false, it means
+ * that it's a dummy relation for which we skipped setting up tlist
+ * expressions and adding EC members in set_append_rel_size(), so
+ * again we have to fail here.
+ */
+ if (rel1_is_simple && !child_rel1->consider_partitionwise_join)
+ {
+ Assert(child_rel1->reloptkind == RELOPT_OTHER_MEMBER_REL);
+ Assert(IS_DUMMY_REL(child_rel1));
+ joinrel->nparts = 0;
+ return;
+ }
+ if (rel2_is_simple && !child_rel2->consider_partitionwise_join)
+ {
+ Assert(child_rel2->reloptkind == RELOPT_OTHER_MEMBER_REL);
+ Assert(IS_DUMMY_REL(child_rel2));
+ joinrel->nparts = 0;
+ return;
+ }
+
+ /* We should never try to join two overlapping sets of rels. */
+ Assert(!bms_overlap(child_rel1->relids, child_rel2->relids));
+ child_joinrelids = bms_union(child_rel1->relids, child_rel2->relids);
+ appinfos = find_appinfos_by_relids(root, child_joinrelids, &nappinfos);
+
+ /*
+ * Construct SpecialJoinInfo from parent join relations's
+ * SpecialJoinInfo.
+ */
+ child_sjinfo = build_child_join_sjinfo(root, parent_sjinfo,
+ child_rel1->relids,
+ child_rel2->relids);
+
+ /*
+ * Construct restrictions applicable to the child join from those
+ * applicable to the parent join.
+ */
+ child_restrictlist =
+ (List *) adjust_appendrel_attrs(root,
+ (Node *) parent_restrictlist,
+ nappinfos, appinfos);
+ pfree(appinfos);
+
+ child_joinrel = joinrel->part_rels[cnt_parts];
+ if (!child_joinrel)
+ {
+ child_joinrel = build_child_join_rel(root, child_rel1, child_rel2,
+ joinrel, child_restrictlist,
+ child_sjinfo,
+ child_sjinfo->jointype);
+ joinrel->part_rels[cnt_parts] = child_joinrel;
+ joinrel->all_partrels = bms_add_members(joinrel->all_partrels,
+ child_joinrel->relids);
+ }
+
+ Assert(bms_equal(child_joinrel->relids, child_joinrelids));
+
+ populate_joinrel_with_paths(root, child_rel1, child_rel2,
+ child_joinrel, child_sjinfo,
+ child_restrictlist);
+ }
+}
OSDN Git Service
