Richard Guo <guofenglinux@gmail.com> writes:
> I see. I didn't notice any real issues with that; I was just flagged
> by the XXX comment there, which raises the question of whether it's
> worth working harder to determine the inputOps.
I was intending to add some code to my nodeSetop patch to see if
both input plan nodes use the same fixed slot type, and if so
pass that rather than NULL to BuildTupleHashTableExt. Perhaps
nodeRecursiveunion could do the same thing (in which case we
probably ought to abstract that out to a subroutine).
regards, tom lane