Robert Haas <robertmhaas@gmail.com> writes:
> So this is about a cross-type join,
> not multiple types within a single partitioning hierarchy, as you
> might also gather from the subject line of this thread.
OK, but I still don't understand why any type conversion is needed
in such a case. The existing join estimators don't try to do that,
for the good and sufficient reasons you and I have already mentioned.
They just apply the given cross-type join operator, and whatever
cross-type selectivity estimator might be associated with it, and
possibly other cross-type operators obtained from the same btree
opfamily.
The minute you get into trying to do any type conversion that is not
mandated by the semantics of the query as written, you're going to
have problems.
regards, tom lane