예제 #1
0
/*
 * have_relevant_joinclause
 *		Detect whether there is a joinclause that involves
 *		the two given relations.
 *
 * Note: the joinclause does not have to be evaluatable with only these two
 * relations.  This is intentional.  For example consider
 *		SELECT * FROM a, b, c WHERE a.x = (b.y + c.z)
 * If a is much larger than the other tables, it may be worthwhile to
 * cross-join b and c and then use an inner indexscan on a.x.  Therefore
 * we should consider this joinclause as reason to join b to c, even though
 * it can't be applied at that join step.
 */
bool
have_relevant_joinclause(PlannerInfo *root,
						 RelOptInfo *rel1, RelOptInfo *rel2)
{
	bool		result = false;
	List	   *joininfo;
	Relids		other_relids;
	ListCell   *l;

	/*
	 * We could scan either relation's joininfo list; may as well use the
	 * shorter one.
	 */
	if (list_length(rel1->joininfo) <= list_length(rel2->joininfo))
	{
		joininfo = rel1->joininfo;
		other_relids = rel2->relids;
	}
	else
	{
		joininfo = rel2->joininfo;
		other_relids = rel1->relids;
	}

	foreach(l, joininfo)
	{
		RestrictInfo *rinfo = (RestrictInfo *) lfirst(l);

		if (bms_overlap(other_relids, rinfo->required_relids))
		{
			result = true;
			break;
		}
	}
예제 #2
0
/*
 * Checks if any of the 'attnums' is a partition key attribute for rel
 *
 * Sets *used_in_expr if any of the 'attnums' is found to be referenced in some
 * partition key expression.  It's possible for a column to be both used
 * directly and as part of an expression; if that happens, *used_in_expr may
 * end up as either true or false.  That's OK for current uses of this
 * function, because *used_in_expr is only used to tailor the error message
 * text.
 */
bool
has_partition_attrs(Relation rel, Bitmapset *attnums, bool *used_in_expr)
{
	PartitionKey key;
	int			partnatts;
	List	   *partexprs;
	ListCell   *partexprs_item;
	int			i;

	if (attnums == NULL || rel->rd_rel->relkind != RELKIND_PARTITIONED_TABLE)
		return false;

	key = RelationGetPartitionKey(rel);
	partnatts = get_partition_natts(key);
	partexprs = get_partition_exprs(key);

	partexprs_item = list_head(partexprs);
	for (i = 0; i < partnatts; i++)
	{
		AttrNumber	partattno = get_partition_col_attnum(key, i);

		if (partattno != 0)
		{
			if (bms_is_member(partattno - FirstLowInvalidHeapAttributeNumber,
							  attnums))
			{
				if (used_in_expr)
					*used_in_expr = false;
				return true;
			}
		}
		else
		{
			/* Arbitrary expression */
			Node	   *expr = (Node *) lfirst(partexprs_item);
			Bitmapset  *expr_attrs = NULL;

			/* Find all attributes referenced */
			pull_varattnos(expr, 1, &expr_attrs);
			partexprs_item = lnext(partexprs_item);

			if (bms_overlap(attnums, expr_attrs))
			{
				if (used_in_expr)
					*used_in_expr = true;
				return true;
			}
		}
	}

	return false;
}
예제 #3
0
파일: hooks.c 프로젝트: akuzm/pg_pathman
static inline bool
allow_star_schema_join(PlannerInfo *root,
					   Path *outer_path,
					   Path *inner_path)
{
	Relids		innerparams = PATH_REQ_OUTER(inner_path);
	Relids		outerrelids = outer_path->parent->relids;

	/*
	 * It's a star-schema case if the outer rel provides some but not all of
	 * the inner rel's parameterization.
	 */
	return (bms_overlap(innerparams, outerrelids) &&
			bms_nonempty_difference(innerparams, outerrelids));
}
예제 #4
0
/*
 * 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 happen
 * when working with outer joins, or with IN or EXISTS clauses that have been
 * turned into joins.
 */
RelOptInfo *
make_join_rel(PlannerInfo *root, RelOptInfo *rel1, RelOptInfo *rel2)
{
	Relids		joinrelids;
	SpecialJoinInfo *sjinfo;
	bool		reversed;
	SpecialJoinInfo sjinfo_data;
	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);

	/* Check validity and determine join type. */
	if (!join_is_legal(root, rel1, rel2, joinrelids,
					   &sjinfo, &reversed))
	{
		/* invalid join path */
		bms_free(joinrelids);
		return NULL;
	}

	/* Swap rels if needed to match the join info. */
	if (reversed)
	{
		RelOptInfo *trel = rel1;

		rel1 = rel2;
		rel2 = trel;
	}

	/*
	 * If it's a plain inner join, then we won't have found anything in
	 * join_info_list.  Make up a SpecialJoinInfo so that selectivity
	 * estimation functions will know what's being joined.
	 */
	if (sjinfo == NULL)
	{
		sjinfo = &sjinfo_data;
		sjinfo->type = T_SpecialJoinInfo;
		sjinfo->min_lefthand = rel1->relids;
		sjinfo->min_righthand = rel2->relids;
		sjinfo->syn_lefthand = rel1->relids;
		sjinfo->syn_righthand = rel2->relids;
		sjinfo->jointype = JOIN_INNER;
		/* we don't bother trying to make the remaining fields valid */
		sjinfo->lhs_strict = false;
		sjinfo->delay_upper_joins = false;
		sjinfo->semi_can_btree = false;
		sjinfo->semi_can_hash = false;
		sjinfo->semi_operators = NIL;
		sjinfo->semi_rhs_exprs = NIL;
	}

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

	/* !!! START: HERE IS THE PART WHICH ADDED FOR PG_HINT_PLAN !!! */
	{
		RowsHint   *rows_hint = NULL;
		int			i;
		RowsHint   *justforme = NULL;
		RowsHint   *domultiply = NULL;

		/* Search for applicable rows hint for this join node */
		for (i = 0; i < current_hint->num_hints[HINT_TYPE_ROWS]; i++)
		{
			rows_hint = current_hint->rows_hints[i];

			/*
			 * Skip this rows_hint if it is invalid from the first or it
			 * doesn't target any join rels.
			 */
			if (!rows_hint->joinrelids ||
				rows_hint->base.state == HINT_STATE_ERROR)
				continue;

			if (bms_equal(joinrelids, rows_hint->joinrelids))
			{
				/*
				 * This joinrel is just the target of this rows_hint, so tweak
				 * rows estimation according to the hint.
				 */
				justforme = rows_hint;
			}
			else if (!(bms_is_subset(rows_hint->joinrelids, rel1->relids) ||
					   bms_is_subset(rows_hint->joinrelids, rel2->relids)) &&
					 bms_is_subset(rows_hint->joinrelids, joinrelids) &&
					 rows_hint->value_type == RVT_MULTI)
			{
				/*
				 * If the rows_hint's target relids is not a subset of both of
				 * component rels and is a subset of this joinrel, ths hint's
				 * targets spread over both component rels. This menas that
				 * this hint has been never applied so far and this joinrel is
				 * the first (and only) chance to fire in current join tree.
				 * Only the multiplication hint has the cumulative nature so we
				 * apply only RVT_MULTI in this way.
				 */
				domultiply = rows_hint;
			}
		}

		if (justforme)
		{
			/*
			 * If a hint just for me is found, no other adjust method is
			 * useles, but this cannot be more than twice becuase this joinrel
			 * is already adjusted by this hint.
			 */
			if (justforme->base.state == HINT_STATE_NOTUSED)
				joinrel->rows = adjust_rows(joinrel->rows, justforme);
		}
		else
		{
			if (domultiply)
			{
				/*
				 * If we have multiple routes up to this joinrel which are not
				 * applicable this hint, this multiply hint will applied more
				 * than twice. But there's no means to know of that,
				 * re-estimate the row number of this joinrel always just
				 * before applying the hint. This is a bit different from
				 * normal planner behavior but it doesn't harm so much.
				 */
				set_joinrel_size_estimates(root, joinrel, rel1, rel2, sjinfo,
										   restrictlist);
				
				joinrel->rows = adjust_rows(joinrel->rows, domultiply);
			}
			
		}
	}
	/* !!! END: HERE IS THE PART WHICH ADDED FOR PG_HINT_PLAN !!! */

	/*
	 * If we've already proven this join is empty, we needn't consider any
	 * more paths for it.
	 */
	if (is_dummy_rel(joinrel))
	{
		bms_free(joinrelids);
		return joinrel;
	}

	/*
	 * Consider paths using each rel as both outer and inner.  Depending on
	 * the join type, a provably empty outer or inner rel might mean the join
	 * is provably empty too; in which case throw away any previously computed
	 * paths and mark the join as dummy.  (We do it this way since it's
	 * conceivable that dummy-ness of a multi-element join might only be
	 * noticeable for certain construction paths.)
	 *
	 * Also, a provably constant-false join restriction typically means that
	 * we can skip evaluating one or both sides of the join.  We do this by
	 * marking the appropriate rel as dummy.  For outer joins, a
	 * constant-false restriction that is pushed down still means the whole
	 * join is dummy, while a non-pushed-down one means that no inner rows
	 * will join so we can treat the inner rel as dummy.
	 *
	 * We need only consider the jointypes that appear in join_info_list, plus
	 * JOIN_INNER.
	 */
	switch (sjinfo->jointype)
	{
		case JOIN_INNER:
			if (is_dummy_rel(rel1) || is_dummy_rel(rel2) ||
				restriction_is_constant_false(restrictlist, false))
			{
				mark_dummy_rel(joinrel);
				break;
			}
			add_paths_to_joinrel(root, joinrel, rel1, rel2,
								 JOIN_INNER, sjinfo,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1,
								 JOIN_INNER, sjinfo,
								 restrictlist);
			break;
		case JOIN_LEFT:
			if (is_dummy_rel(rel1) ||
				restriction_is_constant_false(restrictlist, true))
			{
				mark_dummy_rel(joinrel);
				break;
			}
			if (restriction_is_constant_false(restrictlist, false) &&
				bms_is_subset(rel2->relids, sjinfo->syn_righthand))
				mark_dummy_rel(rel2);
			add_paths_to_joinrel(root, joinrel, rel1, rel2,
								 JOIN_LEFT, sjinfo,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1,
								 JOIN_RIGHT, sjinfo,
								 restrictlist);
			break;
		case JOIN_FULL:
			if ((is_dummy_rel(rel1) && is_dummy_rel(rel2)) ||
				restriction_is_constant_false(restrictlist, true))
			{
				mark_dummy_rel(joinrel);
				break;
			}
			add_paths_to_joinrel(root, joinrel, rel1, rel2,
								 JOIN_FULL, sjinfo,
								 restrictlist);
			add_paths_to_joinrel(root, joinrel, rel2, rel1,
								 JOIN_FULL, sjinfo,
								 restrictlist);

			/*
			 * If there are join quals that aren't mergeable or hashable, we
			 * may not be able to build any valid plan.  Complain here so that
			 * we can give a somewhat-useful error message.  (Since we have no
			 * flexibility of planning for a full join, there's no chance of
			 * succeeding later with another pair of input rels.)
			 */
			if (joinrel->pathlist == NIL)
				ereport(ERROR,
						(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
						 errmsg("FULL JOIN is only supported with merge-joinable or hash-joinable join conditions")));
			break;
		case JOIN_SEMI:

			/*
			 * We might have a normal semijoin, or a case where we don't have
			 * enough rels to do the semijoin but can unique-ify the RHS and
			 * then do an innerjoin (see comments in join_is_legal).  In the
			 * latter case we can't apply JOIN_SEMI joining.
			 */
			if (bms_is_subset(sjinfo->min_lefthand, rel1->relids) &&
				bms_is_subset(sjinfo->min_righthand, rel2->relids))
			{
				if (is_dummy_rel(rel1) || is_dummy_rel(rel2) ||
					restriction_is_constant_false(restrictlist, false))
				{
					mark_dummy_rel(joinrel);
					break;
				}
				add_paths_to_joinrel(root, joinrel, rel1, rel2,
									 JOIN_SEMI, sjinfo,
									 restrictlist);
			}

			/*
			 * If we know how to unique-ify the RHS and one input rel is
			 * exactly the RHS (not a superset) we can consider unique-ifying
			 * it and then doing a regular join.  (The create_unique_path
			 * check here is probably redundant with what join_is_legal did,
			 * but if so the check is cheap because it's cached.  So test
			 * anyway to be sure.)
			 */
			if (bms_equal(sjinfo->syn_righthand, rel2->relids) &&
				create_unique_path(root, rel2, rel2->cheapest_total_path,
								   sjinfo) != NULL)
			{
				if (is_dummy_rel(rel1) || is_dummy_rel(rel2) ||
					restriction_is_constant_false(restrictlist, false))
				{
					mark_dummy_rel(joinrel);
					break;
				}
				add_paths_to_joinrel(root, joinrel, rel1, rel2,
									 JOIN_UNIQUE_INNER, sjinfo,
									 restrictlist);
				add_paths_to_joinrel(root, joinrel, rel2, rel1,
									 JOIN_UNIQUE_OUTER, sjinfo,
									 restrictlist);
			}
			break;
		case JOIN_ANTI:
			if (is_dummy_rel(rel1) ||
				restriction_is_constant_false(restrictlist, true))
			{
				mark_dummy_rel(joinrel);
				break;
			}
			if (restriction_is_constant_false(restrictlist, false) &&
				bms_is_subset(rel2->relids, sjinfo->syn_righthand))
				mark_dummy_rel(rel2);
			add_paths_to_joinrel(root, joinrel, rel1, rel2,
								 JOIN_ANTI, sjinfo,
								 restrictlist);
			break;
		default:
			/* other values not expected here */
			elog(ERROR, "unrecognized join type: %d", (int) sjinfo->jointype);
			break;
	}

	bms_free(joinrelids);

	return joinrel;
}
예제 #5
0
/*
 * make_restrictinfo_internal
 *
 * Common code for the main entry points and the recursive cases.
 */
static RestrictInfo *
make_restrictinfo_internal(Expr *clause,
						   Expr *orclause,
						   bool is_pushed_down,
						   bool outerjoin_delayed,
						   bool pseudoconstant,
						   Index security_level,
						   Relids required_relids,
						   Relids outer_relids,
						   Relids nullable_relids)
{
	RestrictInfo *restrictinfo = makeNode(RestrictInfo);

	restrictinfo->clause = clause;
	restrictinfo->orclause = orclause;
	restrictinfo->is_pushed_down = is_pushed_down;
	restrictinfo->outerjoin_delayed = outerjoin_delayed;
	restrictinfo->pseudoconstant = pseudoconstant;
	restrictinfo->can_join = false; /* may get set below */
	restrictinfo->security_level = security_level;
	restrictinfo->outer_relids = outer_relids;
	restrictinfo->nullable_relids = nullable_relids;

	/*
	 * If it's potentially delayable by lower-level security quals, figure out
	 * whether it's leakproof.  We can skip testing this for level-zero quals,
	 * since they would never get delayed on security grounds anyway.
	 */
	if (security_level > 0)
		restrictinfo->leakproof = !contain_leaked_vars((Node *) clause);
	else
		restrictinfo->leakproof = false;	/* really, "don't know" */

	/*
	 * If it's a binary opclause, set up left/right relids info. In any case
	 * set up the total clause relids info.
	 */
	if (is_opclause(clause) && list_length(((OpExpr *) clause)->args) == 2)
	{
		restrictinfo->left_relids = pull_varnos(get_leftop(clause));
		restrictinfo->right_relids = pull_varnos(get_rightop(clause));

		restrictinfo->clause_relids = bms_union(restrictinfo->left_relids,
												restrictinfo->right_relids);

		/*
		 * Does it look like a normal join clause, i.e., a binary operator
		 * relating expressions that come from distinct relations? If so we
		 * might be able to use it in a join algorithm.  Note that this is a
		 * purely syntactic test that is made regardless of context.
		 */
		if (!bms_is_empty(restrictinfo->left_relids) &&
			!bms_is_empty(restrictinfo->right_relids) &&
			!bms_overlap(restrictinfo->left_relids,
						 restrictinfo->right_relids))
		{
			restrictinfo->can_join = true;
			/* pseudoconstant should certainly not be true */
			Assert(!restrictinfo->pseudoconstant);
		}
	}
	else
	{
		/* Not a binary opclause, so mark left/right relid sets as empty */
		restrictinfo->left_relids = NULL;
		restrictinfo->right_relids = NULL;
		/* and get the total relid set the hard way */
		restrictinfo->clause_relids = pull_varnos((Node *) clause);
	}

	/* required_relids defaults to clause_relids */
	if (required_relids != NULL)
		restrictinfo->required_relids = required_relids;
	else
		restrictinfo->required_relids = restrictinfo->clause_relids;

	/*
	 * Fill in all the cacheable fields with "not yet set" markers. None of
	 * these will be computed until/unless needed.  Note in particular that we
	 * don't mark a binary opclause as mergejoinable or hashjoinable here;
	 * that happens only if it appears in the right context (top level of a
	 * joinclause list).
	 */
	restrictinfo->parent_ec = NULL;

	restrictinfo->eval_cost.startup = -1;
	restrictinfo->norm_selec = -1;
	restrictinfo->outer_selec = -1;

	restrictinfo->mergeopfamilies = NIL;

	restrictinfo->left_ec = NULL;
	restrictinfo->right_ec = NULL;
	restrictinfo->left_em = NULL;
	restrictinfo->right_em = NULL;
	restrictinfo->scansel_cache = NIL;

	restrictinfo->outer_is_left = false;

	restrictinfo->hashjoinoperator = InvalidOid;

	restrictinfo->left_bucketsize = -1;
	restrictinfo->right_bucketsize = -1;
	restrictinfo->left_mcvfreq = -1;
	restrictinfo->right_mcvfreq = -1;

	return restrictinfo;
}