コメントの空白除去などのクリーンアップをした。

[pghintplan/pg_hint_plan.git] / core.c

/*-------------------------------------------------------------------------
 *
 * core.c
 *        Routines copied from PostgreSQL core distribution.
 *
 * src/backend/optimizer/path/allpaths.c
 *     standard_join_search()
 *     set_plain_rel_pathlist()
 *     set_append_rel_pathlist()
 *     accumulate_append_subpath()
 *     set_dummy_rel_pathlist()
 *
 * src/backend/optimizer/path/joinrels.c:
 *     join_search_one_level()
 *     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()
 *
 * Portions Copyright (c) 1996-2011, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 *-------------------------------------------------------------------------
 */

/*
 * 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
 *              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.
 *              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.
 */
RelOptInfo *
standard_join_search(PlannerInfo *root, int levels_needed, List *initial_rels)
{
        int                     lev;
        RelOptInfo *rel;

        /*
         * This function cannot be invoked recursively within any one planning
         * problem, so join_rel_level[] can't be in use already.
         */
        Assert(root->join_rel_level == NULL);

        /*
         * We employ a simple "dynamic programming" algorithm: we first find all
         * ways to build joins of two jointree items, then all ways to build joins
         * of three items (from two-item joins and single items), then four-item
         * joins, and so on until we have considered all ways to join all the
         * items into one rel.
         *
         * root->join_rel_level[j] is a list of all the j-item rels.  Initially we
         * set root->join_rel_level[1] to represent all the single-jointree-item
         * relations.
         */
        root->join_rel_level = (List **) palloc0((levels_needed + 1) * sizeof(List *));

        root->join_rel_level[1] = initial_rels;

        for (lev = 2; lev <= levels_needed; lev++)
        {
                ListCell   *lc;

                /*
                 * Determine all possible pairs of relations to be joined at this
                 * level, and build paths for making each one from every available
                 * pair of lower-level relations.
                 */
                join_search_one_level(root, lev);

                /*
                 * Do cleanup work on each just-processed rel.
                 */
                foreach(lc, root->join_rel_level[lev])
                {
                        rel = (RelOptInfo *) lfirst(lc);

                        /* Find and save the cheapest paths for this rel */
                        set_cheapest(rel);

#ifdef OPTIMIZER_DEBUG
                        debug_print_rel(root, rel);
#endif
                }
        }

        /*
         * We should have a single rel at the final level.
         */
        if (root->join_rel_level[levels_needed] == NIL)
                elog(ERROR, "failed to build any %d-way joins", levels_needed);
        Assert(list_length(root->join_rel_level[levels_needed]) == 1);

        rel = (RelOptInfo *) linitial(root->join_rel_level[levels_needed]);

        root->join_rel_level = NULL;

        return 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.
 */
static void
set_append_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
                                                Index rti, RangeTblEntry *rte)
{
        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));

        /*
         * 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;

                /* 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;

                /* list_copy is important here to avoid sharing list substructure */
                return list_concat(subpaths, list_copy(apath->subpaths));
        }
        else
                return lappend(subpaths, path);
}

/*
 * 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
 *        Consider ways to produce join relations containing exactly 'level'
 *        jointree items.  (This is one step of the dynamic-programming method
 *        embodied in standard_join_search.)  Join rel nodes for each feasible
 *        combination of lower-level rels are created and returned in a list.
 *        Implementation paths are created for each such joinrel, too.
 *
 * level: level of rels we want to make this time
 * root->join_rel_level[j], 1 <= j < level, is a list of rels containing j items
 *
 * The result is returned in root->join_rel_level[level].
 */
void
join_search_one_level(PlannerInfo *root, int level)
{
        List      **joinrels = root->join_rel_level;
        ListCell   *r;
        int                     k;

        Assert(joinrels[level] == NIL);

        /* Set join_cur_level so that new joinrels are added to proper list */
        root->join_cur_level = level;

        /*
         * First, consider left-sided and right-sided plans, in which rels of
         * exactly level-1 member relations are joined against initial relations.
         * 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.
                         *
                         * See also the last-ditch case below.
                         */
                        make_rels_by_clause_joins(root,
                                                                          old_rel,
                                                                          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.
                         */
                        make_rels_by_clauseless_joins(root,
                                                                                  old_rel,
                                                                                  other_rels);
                }
        }

        /*
         * Now, consider "bushy plans" in which relations of k initial rels are
         * joined to relations of level-k initial rels, for 2 <= k <= level-2.
         *
         * We only consider bushy-plan joins for pairs of rels where there is a
         * suitable join clause (or join order restriction), in order to avoid
         * unreasonable growth of planning time.
         */
        for (k = 2;; k++)
        {
                int                     other_level = level - k;

                /*
                 * Since make_join_rel(x, y) handles both x,y and y,x cases, we only
                 * need to go as far as the halfway point.
                 */
                if (k > other_level)
                        break;

                foreach(r, joinrels[k])
                {
                        RelOptInfo *old_rel = (RelOptInfo *) lfirst(r);
                        ListCell   *other_rels;
                        ListCell   *r2;

                        /*
                         * We can ignore clauseless joins here, *except* when they
                         * participate in join-order restrictions --- then we might have
                         * to force a bushy join plan.
                         */
                        if (old_rel->joininfo == NIL && !old_rel->has_eclass_joins &&
                                !has_join_restriction(root, old_rel))
                                continue;

                        if (k == other_level)
                                other_rels = lnext(r);  /* only consider remaining rels */
                        else
                                other_rels = list_head(joinrels[other_level]);

                        for_each_cell(r2, other_rels)
                        {
                                RelOptInfo *new_rel = (RelOptInfo *) lfirst(r2);

                                if (!bms_overlap(old_rel->relids, new_rel->relids))
                                {
                                        /*
                                         * 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.
                                         */
                                        if (have_relevant_joinclause(root, old_rel, new_rel) ||
                                                have_join_order_restriction(root, old_rel, new_rel))
                                        {
                                                (void) make_join_rel(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
         *
         * SELECT * FROM a,b,c WHERE (a.f1 + b.f2 + c.f3) = 0;
         *
         * 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 (joinrels[level] == NIL)
        {
                /*
                 * This loop is just like the first one, except we always call
                 * make_rels_by_clauseless_joins().
                 */
                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);
                }

                /*----------
                 * 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
                 *
                 * SELECT ... FROM t1 WHERE
                 *       x IN (SELECT ... FROM t2,t3 WHERE ...) AND
                 *       y IN (SELECT ... FROM t4,t5 WHERE ...)
                 *
                 * We will flatten this query to a 5-way join problem, but there are
                 * no 4-way joins that join_is_legal() will consider legal.  We have
                 * 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.
                 *----------
                 */
                if (joinrels[level] == NIL && root->join_info_list == NIL)
                        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
 *        that participate in join clauses that 'old_rel' also participates in
 *        (or participate in join-order restrictions with it).
 *        The join rels are returned in root->join_rel_level[join_cur_level].
 *
 * Note: at levels above 2 we will generate the same joined relation in
 * multiple ways --- for example (a join b) join c is the same RelOptInfo as
 * (b join c) join a, though the second case will add a different set of Paths
 * to it.  This is the reason for using the join_rel_level mechanism, which
 * 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
 * rels to be considered for joining
 *
 * 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,
                                                  ListCell *other_rels)
{
        ListCell   *l;

        for_each_cell(l, other_rels)
        {
                RelOptInfo *other_rel = (RelOptInfo *) lfirst(l);

                if (!bms_overlap(old_rel->relids, other_rel->relids) &&
                        (have_relevant_joinclause(root, old_rel, other_rel) ||
                         have_join_order_restriction(root, old_rel, other_rel)))
                {
                        (void) make_join_rel(root, old_rel, other_rel);
                }
        }
}

/*
 * make_rels_by_clauseless_joins
 *        Given a relation 'old_rel' and a list of other relations
 *        'other_rels', create a join relation between 'old_rel' and each
 *        member of 'other_rels' that isn't already included in 'old_rel'.
 *        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
 *
 * 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)
{
        ListCell   *l;

        for_each_cell(l, other_rels)
        {
                RelOptInfo *other_rel = (RelOptInfo *) lfirst(l);

                if (!bms_overlap(other_rel->relids, old_rel->relids))
                {
                        (void) make_join_rel(root, old_rel, other_rel);
                }
        }
}

/*
 * join_is_legal
 *         Determine whether a proposed join is legal given the query's
 *         join order constraints; and if it is, determine the join type.
 *
 * Caller must supply not only the two rels, but the union of their relids.
 * (We could simplify the API by computing joinrelids locally, but this
 * would be redundant work in the normal path through make_join_rel.)
 *
 * 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
 * match the SpecialJoinInfo node.
 */
static bool
join_is_legal(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2,
                          Relids joinrelids,
                          SpecialJoinInfo **sjinfo_p, bool *reversed_p)
{
        SpecialJoinInfo *match_sjinfo;
        bool            reversed;
        bool            unique_ified;
        bool            is_valid_inner;
        ListCell   *l;

        /*
         * Ensure output params are set on failure return.      This is just to
         * suppress uninitialized-variable warnings from overly anal compilers.
         */
        *sjinfo_p = NULL;
        *reversed_p = false;

        /*
         * 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.
         */
        match_sjinfo = NULL;
        reversed = false;
        unique_ified = false;
        is_valid_inner = true;

        foreach(l, root->join_info_list)
        {
                SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l);

                /*
                 * This special join is not relevant unless its RHS overlaps the
                 * proposed join.  (Check this first as a fast path for dismissing
                 * most irrelevant SJs quickly.)
                 */
                if (!bms_overlap(sjinfo->min_righthand, joinrelids))
                        continue;

                /*
                 * Also, not relevant if proposed join is fully contained within RHS
                 * (ie, we're still building up the RHS).
                 */
                if (bms_is_subset(joinrelids, sjinfo->min_righthand))
                        continue;

                /*
                 * Also, not relevant if SJ is already done within either input.
                 */
                if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) &&
                        bms_is_subset(sjinfo->min_righthand, rel1->relids))
                        continue;
                if (bms_is_subset(sjinfo->min_lefthand, rel2->relids) &&
                        bms_is_subset(sjinfo->min_righthand, rel2->relids))
                        continue;

                /*
                 * If it's a semijoin and we already joined the RHS to any other rels
                 * within either input, then we must have unique-ified the RHS at that
                 * point (see below).  Therefore the semijoin is no longer relevant in
                 * this join path.
                 */
                if (sjinfo->jointype == JOIN_SEMI)
                {
                        if (bms_is_subset(sjinfo->syn_righthand, rel1->relids) &&
                                !bms_equal(sjinfo->syn_righthand, rel1->relids))
                                continue;
                        if (bms_is_subset(sjinfo->syn_righthand, rel2->relids) &&
                                !bms_equal(sjinfo->syn_righthand, rel2->relids))
                                continue;
                }

                /*
                 * 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?)
                 */
                if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) &&
                        bms_is_subset(sjinfo->min_righthand, rel2->relids))
                {
                        if (match_sjinfo)
                                return false;   /* invalid join path */
                        match_sjinfo = sjinfo;
                        reversed = false;
                }
                else if (bms_is_subset(sjinfo->min_lefthand, rel2->relids) &&
                                 bms_is_subset(sjinfo->min_righthand, rel1->relids))
                {
                        if (match_sjinfo)
                                return false;   /* invalid join path */
                        match_sjinfo = sjinfo;
                        reversed = true;
                }
                else if (sjinfo->jointype == JOIN_SEMI &&
                                 bms_equal(sjinfo->syn_righthand, rel2->relids) &&
                                 create_unique_path(root, rel2, rel2->cheapest_total_path,
                                                                        sjinfo) != NULL)
                {
                        /*----------
                         * For a semijoin, we can join the RHS to anything else by
                         * unique-ifying the RHS (if the RHS can be unique-ified).
                         * We will only get here if we have the full RHS but less
                         * than min_lefthand on the LHS.
                         *
                         * The reason to consider such a join path is exemplified by
                         *      SELECT ... FROM a,b WHERE (a.x,b.y) IN (SELECT c1,c2 FROM c)
                         * If we insist on doing this as a semijoin we will first have
                         * to form the cartesian product of A*B.  But if we unique-ify
                         * C then the semijoin becomes a plain innerjoin and we can join
                         * in any order, eg C to A and then to B.  When C is much smaller
                         * than A and B this can be a huge win.  So we allow C to be
                         * joined to just A or just B here, and then make_join_rel has
                         * to handle the case properly.
                         *
                         * Note that actually we'll allow unique-ified C to be joined to
                         * some other relation D here, too.  That is legal, if usually not
                         * very sane, and this routine is only concerned with legality not
                         * with whether the join is good strategy.
                         *----------
                         */
                        if (match_sjinfo)
                                return false;   /* invalid join path */
                        match_sjinfo = sjinfo;
                        reversed = false;
                        unique_ified = true;
                }
                else if (sjinfo->jointype == JOIN_SEMI &&
                                 bms_equal(sjinfo->syn_righthand, rel1->relids) &&
                                 create_unique_path(root, rel1, rel1->cheapest_total_path,
                                                                        sjinfo) != NULL)
                {
                        /* Reversed semijoin case */
                        if (match_sjinfo)
                                return false;   /* invalid join path */
                        match_sjinfo = sjinfo;
                        reversed = true;
                        unique_ified = true;
                }
                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.
                         *----------
                         */
                        if (sjinfo->jointype != JOIN_SEMI &&
                                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;
                }
        }

        /*
         * 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).
         */
        if (!is_valid_inner &&
                (match_sjinfo == NULL || unique_ified))
                return false;                   /* invalid join path */

        /* 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).
 *
 * 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
 * expensive has_legal_joinclause test.)
 */
static bool
has_join_restriction(PlannerInfo *root, RelOptInfo *rel)
{
        ListCell   *l;

        foreach(l, root->join_info_list)
        {
                SpecialJoinInfo *sjinfo = (SpecialJoinInfo *) lfirst(l);

                /* ignore full joins --- other mechanisms preserve their ordering */
                if (sjinfo->jointype == JOIN_FULL)
                        continue;

                /* ignore if SJ is already contained in rel */
                if (bms_is_subset(sjinfo->min_lefthand, rel->relids) &&
                        bms_is_subset(sjinfo->min_righthand, rel->relids))
                        continue;

                /* restricted if it overlaps LHS or RHS, but doesn't contain SJ */
                if (bms_overlap(sjinfo->min_lefthand, rel->relids) ||
                        bms_overlap(sjinfo->min_righthand, rel->relids))
                        return true;
        }

        return false;
}

/*
 * 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
 * will generally have thrown away anything that's ANDed with it.  In outer
 * join situations this will leave us computing cartesian products only to
 * 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.
 */
static bool
restriction_is_constant_false(List *restrictlist, bool only_pushed_down)
{
        ListCell   *lc;

        /*
         * Despite the above comment, the restriction list we see here might
         * possibly have other members besides the FALSE constant, since other
         * quals could get "pushed down" to the outer join level.  So we check
         * each member of the list.
         */
        foreach(lc, restrictlist)
        {
                RestrictInfo *rinfo = (RestrictInfo *) lfirst(lc);

                Assert(IsA(rinfo, RestrictInfo));
                if (only_pushed_down && !rinfo->is_pushed_down)
                        continue;

                if (rinfo->clause && IsA(rinfo->clause, Const))
                {
                        Const      *con = (Const *) rinfo->clause;

                        /* constant NULL is as good as constant FALSE for our purposes */
                        if (con->constisnull)
                                return true;
                        if (!DatumGetBool(con->constvalue))
                                return true;
                }
        }
        return false;
}