On Jan 18, Tom Lane modulated:
> Karl Czajkowski <karlcz@isi.edu> writes:
> > Is there a correctness hazard with pretending our function is
> > IMMUTABLE, even though we will change the underlying config parameter
> > in the same connection?
>
> You could probably get away with that if you never ever use prepared
> queries (beware that almost anything in plpgsql is a prepared query).
> It's a trick that's likely to bite you eventually though.
>
That sounds unnerving. I think I need to play it safe. :-/
Does the plan cache disappear with each connection/backend process?
Or is there also a risk of plans being shared between backends?
Would it be invasive or a small hack to have something like
"transaction-immutable" which can be precomputed during planning, like
immutable, but then must discard those plans at the end of the
transaction...?
karl