My question can be demonstrated with the below example:
create table m1(a int, b int);
explain (costs off) select (select count(*) filter (where true) from m1 t1)
from m1 t2 where t2.b % 2 = 1;
QUERY PLAN
---------------------------------
Seq Scan on m1 t2
Filter: ((b % 2) = 1)
InitPlan 1 (returns $0)
-> Aggregate
-> Seq Scan on m1 t1
(5 rows)
The above is good to me. The aggregate is run in the subPlan/InitPlan.
explain (costs off) select (select count(*) filter (
where t2.b = 1) from m1 t1)
from m1 t2 where t2.b % 2 = 1;
QUERY PLAN
-------------------------------
Aggregate
-> Seq Scan on m1 t2
Filter: ((b % 2) = 1)
SubPlan 1
-> Seq Scan on m1 t1
(5 rows)
This one is too confusing to me since the Aggregate happens
on t2 rather than t1. What happens here? Would this query
generate 1 row all the time like SELECT aggfunc(a) FROM t?
--