"Paragon Corporation" <lr@pcorp.us> writes:
>> Yes. The FDW author is not expected to implement those clauses; where do
>> you see documentation saying otherwise?
> I didn't. I was copying the postgres_fdw and noticed it implemented them
> directly.
Ah. I believe that's mostly just an efficiency hack (and a fairly trivial
one too, at least in postgres_fdw's case). I suppose there might be an
argument that if we don't exclude tables at this stage, we could get a
parse-time failure from features not supported on the local server, which
would be unexpected if the user had told us to skip such tables. But that
seems a bit implausible.
I'll go add a comment about this though, to clarify matters for the
next person.
regards, tom lane
