joinrels.c

PostgreSQL 8.1.4的源码 适用于Linux下的开源数据库系统

	List	   *result = NIL;
	ListCell   *i;

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

		if (!bms_overlap(other_rel->relids, old_rel->relids))
		{
			RelOptInfo *jrel;

			jrel = make_join_rel(root, old_rel, other_rel, JOIN_INNER);

			/*
			 * As long as given other_rels are distinct, don't need to test to
			 * see if jrel is already part of output list.
			 */
			if (jrel)
				result = lcons(jrel, result);
		}
	}

	return result;
}


/*
 * is_inside_IN
 *		Detect whether the specified relation is inside an IN (sub-SELECT).
 *
 * Note that we are actually only interested in rels that have been pulled up
 * out of an IN, so the routine name is a slight misnomer.
 */
static bool
is_inside_IN(PlannerInfo *root, RelOptInfo *rel)
{
	ListCell   *l;

	foreach(l, root->in_info_list)
	{
		InClauseInfo *ininfo = (InClauseInfo *) lfirst(l);

		if (bms_is_subset(rel->relids, ininfo->righthand))
			return true;
	}
	return false;
}


/*
 * make_jointree_rel
 *		Find or build a RelOptInfo join rel representing a specific
 *		jointree item.	For JoinExprs, we only consider the construction
 *		path that corresponds exactly to what the user wrote.
 */
RelOptInfo *
make_jointree_rel(PlannerInfo *root, Node *jtnode)
{
	if (IsA(jtnode, RangeTblRef))
	{
		int			varno = ((RangeTblRef *) jtnode)->rtindex;

		return find_base_rel(root, varno);
	}
	else if (IsA(jtnode, FromExpr))
	{
		FromExpr   *f = (FromExpr *) jtnode;

		/* Recurse back to multi-way-join planner */
		return make_fromexpr_rel(root, f);
	}
	else if (IsA(jtnode, JoinExpr))
	{
		JoinExpr   *j = (JoinExpr *) jtnode;
		RelOptInfo *rel,
				   *lrel,
				   *rrel;

		/* Recurse */
		lrel = make_jointree_rel(root, j->larg);
		rrel = make_jointree_rel(root, j->rarg);

		/* Make this join rel */
		rel = make_join_rel(root, lrel, rrel, j->jointype);

		if (rel == NULL)		/* oops */
			elog(ERROR, "invalid join order");

		/*
		 * Since we are only going to consider this one way to do it, we're
		 * done generating Paths for this joinrel and can now select the
		 * cheapest.  In fact we *must* do so now, since next level up will
		 * need it!
		 */
		set_cheapest(rel);

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

		return rel;
	}
	else
		elog(ERROR, "unrecognized node type: %d",
			 (int) nodeTag(jtnode));
	return NULL;				/* keep compiler quiet */
}


/*
 * make_join_rel
 *	   Find or create a join RelOptInfo that represents the join of
 *	   the two given rels, and add to it path information for paths
 *	   created with the two rels as outer and inner rel.
 *	   (The join rel may already contain paths generated from other
 *	   pairs of rels that add up to the same set of base rels.)
 *
 * NB: will return NULL if attempted join is not valid.  This can only
 * happen when working with IN clauses that have been turned into joins.
 */
RelOptInfo *
make_join_rel(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2,
			  JoinType jointype)
{
	Relids		joinrelids;
	RelOptInfo *joinrel;
	List	   *restrictlist;

	/* We should never try to join two overlapping sets of rels. */
	Assert(!bms_overlap(rel1->relids, rel2->relids));

	/* Construct Relids set that identifies the joinrel. */
	joinrelids = bms_union(rel1->relids, rel2->relids);

	/*
	 * If we are implementing IN clauses as joins, there are some joins that
	 * are illegal.  Check to see if the proposed join is trouble. We can skip
	 * the work if looking at an outer join, however, because only top-level
	 * joins might be affected.
	 */
	if (jointype == JOIN_INNER)
	{
		ListCell   *l;

		foreach(l, root->in_info_list)
		{
			InClauseInfo *ininfo = (InClauseInfo *) lfirst(l);

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

			/*
			 * If we are still building the IN clause's RHS, then this IN
			 * clause isn't relevant yet.
			 */
			if (bms_is_subset(joinrelids, ininfo->righthand))
				continue;

			/*
			 * Cannot join if proposed join contains rels not in the RHS *and*
			 * contains only part of the RHS.  We must build the complete RHS
			 * (subselect's join) before it can be joined to rels outside the
			 * subselect.
			 */
			if (!bms_is_subset(ininfo->righthand, joinrelids))
			{
				bms_free(joinrelids);
				return NULL;
			}

			/*
			 * At this point we are considering a join of the IN's RHS to some
			 * other rel(s).
			 *
			 * If we already joined IN's RHS to any other rels in either input
			 * path, then this join is not constrained (the necessary work was
			 * done at the lower level where that join occurred).
			 */
			if (bms_is_subset(ininfo->righthand, rel1->relids) &&
				!bms_equal(ininfo->righthand, rel1->relids))
				continue;
			if (bms_is_subset(ininfo->righthand, rel2->relids) &&
				!bms_equal(ininfo->righthand, rel2->relids))
				continue;

			/*
			 * JOIN_IN technique will work if outerrel includes LHS and
			 * innerrel is exactly RHS; conversely JOIN_REVERSE_IN handles
			 * RHS/LHS.
			 *
			 * JOIN_UNIQUE_OUTER will work if outerrel is exactly RHS;
			 * conversely JOIN_UNIQUE_INNER will work if innerrel is exactly
			 * RHS.
			 *
			 * But none of these will work if we already found another IN that
			 * needs to trigger here.
			 */
			if (jointype != JOIN_INNER)
			{
				bms_free(joinrelids);
				return NULL;
			}
			if (bms_is_subset(ininfo->lefthand, rel1->relids) &&
				bms_equal(ininfo->righthand, rel2->relids))
				jointype = JOIN_IN;
			else if (bms_is_subset(ininfo->lefthand, rel2->relids) &&
					 bms_equal(ininfo->righthand, rel1->relids))
				jointype = JOIN_REVERSE_IN;
			else if (bms_equal(ininfo->righthand, rel1->relids))
				jointype = JOIN_UNIQUE_OUTER;
			else if (bms_equal(ininfo->righthand, rel2->relids))
				jointype = JOIN_UNIQUE_INNER;
			else
			{
				/* invalid join path */
				bms_free(joinrelids);
				return NULL;
			}
		}
	}

	/*
	 * Find or build the join RelOptInfo, and compute the restrictlist that
	 * goes with this particular joining.
	 */
	joinrel = build_join_rel(root, joinrelids, rel1, rel2, jointype,
							 &restrictlist);

	/*
	 * Consider paths using each rel as both outer and inner.
	 */
	switch (jointype)
	{
		case JOIN_INNER:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_INNER,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_INNER,
								 restrictlist);
			break;
		case JOIN_LEFT:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_LEFT,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_RIGHT,
								 restrictlist);
			break;
		case JOIN_FULL:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_FULL,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_FULL,
								 restrictlist);
			break;
		case JOIN_RIGHT:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_RIGHT,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_LEFT,
								 restrictlist);
			break;
		case JOIN_IN:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_IN,
								 restrictlist);
			/* REVERSE_IN isn't supported by joinpath.c */
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_UNIQUE_INNER,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_UNIQUE_OUTER,
								 restrictlist);
			break;
		case JOIN_REVERSE_IN:
			/* REVERSE_IN isn't supported by joinpath.c */
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_IN,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_UNIQUE_OUTER,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_UNIQUE_INNER,
								 restrictlist);
			break;
		case JOIN_UNIQUE_OUTER:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_UNIQUE_OUTER,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_UNIQUE_INNER,
								 restrictlist);
			break;
		case JOIN_UNIQUE_INNER:
			add_paths_to_joinrel(root, joinrel, rel1, rel2, JOIN_UNIQUE_INNER,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1, JOIN_UNIQUE_OUTER,
								 restrictlist);
			break;
		default:
			elog(ERROR, "unrecognized join type: %d",
				 (int) jointype);
			break;
	}

	bms_free(joinrelids);

	return joinrel;
}