When pulling up a subquery that is under an outer join, if the
subquery's target list contains a strict expression that uses a
subquery variable, it's okay to pull up the expression without
wrapping it in a PlaceHolderVar: if the subquery variable is forced to
NULL by the outer join, the expression result will come out as NULL
too.
If the strict expression does not contain any subquery variables, the
current code always wraps it in a PlaceHolderVar. While this is not
incorrect, the analysis could be tighter: if the strict expression
contains any variables of rels that are under the same lowest nulling
outer join as the subquery, we can also avoid wrapping it. This is
safe because if the subquery variable is forced to NULL by the outer
join, the variables of rels that are under the same lowest nulling
outer join will also be forced to NULL, resulting in the expression
evaluating to NULL as well. Therefore, it's not necessary to force
the expression to be evaluated below the outer join. It could be
beneficial to get rid of such PHVs because they could imply lateral
dependencies, which force us to resort to nestloop joins.
This patch checks if the lateral references in the strict expression
contain any variables of rels under the same lowest nulling outer join
as the subquery, and avoids wrapping the expression in that case.
This is fundamentally a generalization of the optimizations for bare
Vars and PHVs introduced in commit
f64ec81a8.
No backpatch as this could result in plan changes.
Author: Richard Guo
Discussion: https://postgr.es/m/CAMbWs4_ENtfRdLaM_bXAxiKRYO7DmwDBDG4_2=VTDi0mJP-jAw@mail.gmail.com
{
/*
* If the node contains Var(s) or PlaceHolderVar(s) of the
- * subquery being pulled up, and does not contain any
- * non-strict constructs, then instead of adding a PHV on top
- * we can add the required nullingrels to those Vars/PHVs.
- * (This is fundamentally a generalization of the above cases
- * for bare Vars and PHVs.)
+ * subquery being pulled up, or of rels that are under the
+ * same lowest nulling outer join as the subquery, and does
+ * not contain any non-strict constructs, then instead of
+ * adding a PHV on top we can add the required nullingrels to
+ * those Vars/PHVs. (This is fundamentally a generalization
+ * of the above cases for bare Vars and PHVs.)
*
* This test is somewhat expensive, but it avoids pessimizing
* the plan in cases where the nullingrels get removed again
*
* Note that we don't force wrapping of expressions containing
* lateral references, so long as they also contain Vars/PHVs
- * of the subquery. This is okay because of the restriction
- * to strict constructs: if the subquery's Vars/PHVs have been
- * forced to NULL by an outer join then the end result of the
- * expression will be NULL too, regardless of the lateral
- * references. So it's not necessary to force the expression
- * to be evaluated below the outer join. This can be a very
- * valuable optimization, because it may allow us to avoid
- * using a nested loop to pass the lateral reference down.
+ * of the subquery, or of rels that are under the same lowest
+ * nulling outer join as the subquery. This is okay because
+ * of the restriction to strict constructs: if those Vars/PHVs
+ * have been forced to NULL by an outer join then the end
+ * result of the expression will be NULL too, regardless of
+ * the lateral references. So it's not necessary to force the
+ * expression to be evaluated below the outer join. This can
+ * be a very valuable optimization, because it may allow us to
+ * avoid using a nested loop to pass the lateral reference
+ * down.
*
* This analysis could be tighter: in particular, a non-strict
* construct hidden within a lower-level PlaceHolderVar is not
* membership of the node, but if it's non-lateral then any
* level-zero var must belong to the subquery.
*/
- if ((rcon->target_rte->lateral ?
- bms_overlap(pull_varnos(rcon->root, newnode),
- rcon->relids) :
- contain_vars_of_level(newnode, 0)) &&
+ bool contain_nullable_vars = false;
+
+ if (!rcon->target_rte->lateral)
+ {
+ if (contain_vars_of_level(newnode, 0))
+ contain_nullable_vars = true;
+ }
+ else
+ {
+ Relids all_varnos;
+
+ all_varnos = pull_varnos(rcon->root, newnode);
+ if (bms_overlap(all_varnos, rcon->relids))
+ contain_nullable_vars = true;
+ else
+ {
+ nullingrel_info *nullinfo = rcon->nullinfo;
+ int varno;
+
+ varno = -1;
+ while ((varno = bms_next_member(all_varnos, varno)) >= 0)
+ {
+ Assert(varno > 0 && varno <= nullinfo->rtlength);
+ if (bms_is_subset(nullinfo->nullingrels[rcon->varno],
+ nullinfo->nullingrels[varno]))
+ {
+ contain_nullable_vars = true;
+ break;
+ }
+ }
+ }
+ }
+
+ if (contain_nullable_vars &&
!contain_nonstrict_functions(newnode))
{
/* No wrap needed */
4567890123456789 | 9135780246913578
(11 rows)
+-- strict expressions containing variables of rels under the same lowest
+-- nulling outer join can escape being wrapped
+explain (verbose, costs off)
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 inner join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+ QUERY PLAN
+--------------------------------------------------------
+ Sort
+ Output: t1.q1, ((t2.q1 + 1))
+ Sort Key: t1.q1, ((t2.q1 + 1))
+ -> Hash Right Join
+ Output: t1.q1, (t2.q1 + 1)
+ Hash Cond: (t2.q2 = t1.q2)
+ -> Hash Join
+ Output: t2.q1, t2.q2
+ Hash Cond: (t2.q2 = t3.q2)
+ -> Seq Scan on public.int8_tbl t2
+ Output: t2.q1, t2.q2
+ -> Hash
+ Output: t3.q2
+ -> Seq Scan on public.int8_tbl t3
+ Output: t3.q2
+ -> Hash
+ Output: t1.q1, t1.q2
+ -> Seq Scan on public.int8_tbl t1
+ Output: t1.q1, t1.q2
+(19 rows)
+
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 inner join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+ q1 | x
+------------------+------------------
+ 123 | 124
+ 123 | 124
+ 123 | 124
+ 123 | 4567890123456790
+ 123 | 4567890123456790
+ 4567890123456789 | 124
+ 4567890123456789 | 124
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+(11 rows)
+
+-- otherwise we need to wrap the strict expressions
+explain (verbose, costs off)
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 left join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+ QUERY PLAN
+--------------------------------------------------
+ Sort
+ Output: t1.q1, ((t2.q1 + 1))
+ Sort Key: t1.q1, ((t2.q1 + 1))
+ -> Hash Right Join
+ Output: t1.q1, ((t2.q1 + 1))
+ Hash Cond: (t2.q2 = t1.q2)
+ -> Nested Loop Left Join
+ Output: t2.q2, ((t2.q1 + 1))
+ -> Seq Scan on public.int8_tbl t2
+ Output: t2.q1, t2.q2
+ -> Seq Scan on public.int8_tbl t3
+ Output: t3.q2, (t2.q1 + 1)
+ Filter: (t2.q2 = t3.q2)
+ -> Hash
+ Output: t1.q1, t1.q2
+ -> Seq Scan on public.int8_tbl t1
+ Output: t1.q1, t1.q2
+(17 rows)
+
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 left join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+ q1 | x
+------------------+------------------
+ 123 | 124
+ 123 | 124
+ 123 | 124
+ 123 | 4567890123456790
+ 123 | 4567890123456790
+ 4567890123456789 | 124
+ 4567890123456789 | 124
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+ 4567890123456789 | 4567890123456790
+(11 rows)
+
-- lateral references for simple Vars can escape being wrapped if the
-- referenced rel is under the same lowest nulling outer join
explain (verbose, costs off)
on t1.q2 = t2.q2
order by 1, 2;
+-- strict expressions containing variables of rels under the same lowest
+-- nulling outer join can escape being wrapped
+explain (verbose, costs off)
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 inner join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 inner join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+
+-- otherwise we need to wrap the strict expressions
+explain (verbose, costs off)
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 left join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+
+select t1.q1, x from
+ int8_tbl t1 left join
+ (int8_tbl t2 left join
+ lateral (select t2.q1+1 as x, * from int8_tbl t3) t3 on t2.q2 = t3.q2)
+ on t1.q2 = t2.q2
+order by 1, 2;
+
-- lateral references for simple Vars can escape being wrapped if the
-- referenced rel is under the same lowest nulling outer join
explain (verbose, costs off)
projects / postgresql.git / commitdiff