Multirow subselect as parameter to `execute using`

Question

The multirow subselect will be used in the right hand side of the in operator in the where clause:

create table t (a integer);
insert into t (a) values (1), (9);

drop function if exists f();

create function f()
returns void as $$
begin
execute '
    select a
    from t
    where a in $1
' using (select 1 union select 2);
end;$$
language plpgsql;

select f();

ERROR:  more than one row returned by a subquery used as an expression
CONTEXT:  SQL statement "SELECT (select 1 union select 2)"
PL/pgSQL function "f" line 3 at EXECUTE statement

How to achieve what the above function would if it worked?

Answer 1

I see nothing in your question that couldn't more easily be solved with:

SELECT a
FROM   t
JOIN  (VALUES (1), (2)) x(a) USING (a); -- any query returning multiple int

Can you clarify the necessity in your example?

---

As a proof of concept, this would be simpler / faster:

CREATE OR REPLACE FUNCTION x.f1()
  RETURNS SETOF integer AS
$BODY$
BEGIN
RETURN QUERY EXECUTE '
    SELECT a
    FROM t
    WHERE a = ANY($1)'
USING ARRAY(VALUES (1), (2)); -- any query here
END;
$BODY$
LANGUAGE plpgsql;

---

Performance of IN () and = ANY()

Your observation is spot on. And there is reason for it. Try:

EXPLAIN ANALYZE SELECT * FROM tbl WHERE id IN (1,2,3);

The query plan will reveal:

Index Cond: (id = ANY ('{1,2,3}'::integer[]))

PostgreSQL transforms the id IN (..) construct into id = ANY(..) internally. These two perform identically - except for a negligible overhead penalty.

My code is only faster in building the statement - exactly as you diagnosed.

Answer 2

create function f()
returns setof integer as $$
begin
return query execute format('
    select a
    from t
    where a in %s
', replace(replace(array(select 1 union select 2)::text, '{', '('), '}', ')'));
end$$
language plpgsql;