> How do you find the ANY(ARRAY(...)) version performs with say 10
> million records in the array?
I've tested with a subquery which returns about 20 million different
rows. In this case IN(...) is about 5 times faster than ANY(ARRAY(...))
for me. The exact numbers are:
IN(...): about 22 seconds
ANY(ARRAY(...)): about 115 seconds
However, estimated query costs seem to be always correct. So shouldn't
it be quite easy for the planner to create query plans for both the
ANY(ARRAY(...)) and the IN(...) version and then just use the plan where
costs are cheaper?