Teach planner to transform "x IS [NOT] DISTINCT FROM NULL" to a NullTest

In the spirit of 8d19d0e13, this patch teaches the planner about the
principle that NullTest with !argisrow is fully equivalent to SQL's IS
[NOT] DISTINCT FROM NULL.

The parser already performs this transformation for literal NULLs.
However, a DistinctExpr expression with one input evaluating to NULL
during planning (e.g., via const-folding of "1 + NULL" or parameter
substitution in custom plans) currently remains as a DistinctExpr
node.

This patch closes the gap for const-folded NULLs.  It specifically
targets the case where one input is a constant NULL and the other is a
nullable non-constant expression.  (If the other input were otherwise,
the DistinctExpr node would have already been simplified to a constant
TRUE or FALSE.)

This transformation can be beneficial because NullTest is much more
amenable to optimization than DistinctExpr, since the planner knows a
good deal about the former and next to nothing about the latter.

Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Discussion: https://postgr.es/m/CAMbWs49BMAOWvkdSHxpUDnniqJcEcGq3_8dd_5wTR4xrQY8urA@mail.gmail.com
This commit is contained in:
Richard Guo 2026-02-10 10:19:25 +09:00
parent 0aaf0de7fe
commit f41ab51573
3 changed files with 118 additions and 2 deletions

View file

@ -2837,6 +2837,30 @@ eval_const_expressions_mutator(Node *node,
return eval_const_expressions_mutator(negate_clause((Node *) eqexpr),
context);
}
else if (has_null_input)
{
/*
* One input is a nullable non-constant expression, and
* the other is an explicit NULL constant. We can
* transform this to a NullTest with !argisrow, which is
* much more amenable to optimization.
*/
NullTest *nt = makeNode(NullTest);
nt->arg = (Expr *) (IsA(linitial(args), Const) ?
lsecond(args) : linitial(args));
nt->nulltesttype = IS_NOT_NULL;
/*
* argisrow = false is correct whether or not arg is
* composite
*/
nt->argisrow = false;
nt->location = expr->location;
return eval_const_expressions_mutator((Node *) nt, context);
}
/*
* The expression cannot be simplified any further, so build

View file

@ -633,7 +633,7 @@ SELECT * FROM pred_tab WHERE (a::oid) IS NULL;
DROP TABLE pred_tab;
--
-- Test optimization of IS [NOT] DISTINCT FROM on non-nullable inputs
-- Test optimization of IS [NOT] DISTINCT FROM
--
CREATE TYPE dist_row_t AS (a int, b int);
CREATE TABLE dist_tab (id int, val_nn int NOT NULL, val_null int, row_nn dist_row_t NOT NULL);
@ -766,6 +766,73 @@ SELECT * FROM dist_tab t1 JOIN dist_tab t2 ON t1.val_nn IS NOT DISTINCT FROM t2.
(3 rows)
RESET enable_nestloop;
-- Ensure that the predicate is converted to IS NOT NULL
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE val_null IS DISTINCT FROM NULL::INT;
QUERY PLAN
----------------------------------
Seq Scan on dist_tab
Filter: (val_null IS NOT NULL)
(2 rows)
SELECT id FROM dist_tab WHERE val_null IS DISTINCT FROM NULL::INT;
id
----
1
3
(2 rows)
-- Ensure that the predicate is converted to IS NULL
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE val_null IS NOT DISTINCT FROM NULL::INT;
QUERY PLAN
------------------------------
Seq Scan on dist_tab
Filter: (val_null IS NULL)
(2 rows)
SELECT id FROM dist_tab WHERE val_null IS NOT DISTINCT FROM NULL::INT;
id
----
2
(1 row)
-- Safety check for rowtypes
-- The predicate is converted to IS NOT NULL, and get_rule_expr prints it as IS
-- DISTINCT FROM because argisrow is false, indicating that we're applying a
-- scalar test
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS DISTINCT FROM NULL::RECORD;
QUERY PLAN
-----------------------------------------------------------
Seq Scan on dist_tab
Filter: (ROW(val_null, val_null) IS DISTINCT FROM NULL)
(2 rows)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS DISTINCT FROM NULL::RECORD;
id
----
1
2
3
(3 rows)
-- The predicate is converted to IS NULL, and get_rule_expr prints it as IS NOT
-- DISTINCT FROM because argisrow is false, indicating that we're applying a
-- scalar test
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS NOT DISTINCT FROM NULL::RECORD;
QUERY PLAN
---------------------------------------------------------------
Seq Scan on dist_tab
Filter: (ROW(val_null, val_null) IS NOT DISTINCT FROM NULL)
(2 rows)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS NOT DISTINCT FROM NULL::RECORD;
id
----
(0 rows)
DROP TABLE dist_tab;
DROP TYPE dist_row_t;
--

View file

@ -310,7 +310,7 @@ SELECT * FROM pred_tab WHERE (a::oid) IS NULL;
DROP TABLE pred_tab;
--
-- Test optimization of IS [NOT] DISTINCT FROM on non-nullable inputs
-- Test optimization of IS [NOT] DISTINCT FROM
--
CREATE TYPE dist_row_t AS (a int, b int);
@ -367,6 +367,31 @@ SELECT * FROM dist_tab t1 JOIN dist_tab t2 ON t1.val_nn IS NOT DISTINCT FROM t2.
SELECT * FROM dist_tab t1 JOIN dist_tab t2 ON t1.val_nn IS NOT DISTINCT FROM t2.val_nn;
RESET enable_nestloop;
-- Ensure that the predicate is converted to IS NOT NULL
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE val_null IS DISTINCT FROM NULL::INT;
SELECT id FROM dist_tab WHERE val_null IS DISTINCT FROM NULL::INT;
-- Ensure that the predicate is converted to IS NULL
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE val_null IS NOT DISTINCT FROM NULL::INT;
SELECT id FROM dist_tab WHERE val_null IS NOT DISTINCT FROM NULL::INT;
-- Safety check for rowtypes
-- The predicate is converted to IS NOT NULL, and get_rule_expr prints it as IS
-- DISTINCT FROM because argisrow is false, indicating that we're applying a
-- scalar test
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS DISTINCT FROM NULL::RECORD;
SELECT id FROM dist_tab WHERE (val_null, val_null) IS DISTINCT FROM NULL::RECORD;
-- The predicate is converted to IS NULL, and get_rule_expr prints it as IS NOT
-- DISTINCT FROM because argisrow is false, indicating that we're applying a
-- scalar test
EXPLAIN (COSTS OFF)
SELECT id FROM dist_tab WHERE (val_null, val_null) IS NOT DISTINCT FROM NULL::RECORD;
SELECT id FROM dist_tab WHERE (val_null, val_null) IS NOT DISTINCT FROM NULL::RECORD;
DROP TABLE dist_tab;
DROP TYPE dist_row_t;