Bruno Wolff III <bruno@wolff.to> writes:
> Thinking about the case for NULLs some more, I am wondering if you are
> going to treat aggregates with strict state functions different than
> those that don't?
We only intend this to support MAX and MIN. If you're inventing an
aggregate that doesn't play exactly by those rules, you don't get to
take advantage of the optimization.
regards, tom lane