Re: Push down time-related SQLValue functions to foreign server

Alexander Pyhalov <a(dot)pyhalov(at)postgrespro(dot)ru> writes:
> Tom Lane писал 2022-01-18 19:53:
>> However, before we proceed any further with this patch, I think we
>> really ought to stop and think about the question I raised last
>> night: why are we building a one-off feature for SQLValueFunction?
>> Wouldn't the same parameter-substitution mechanism work for *any*
>> stable expression that doesn't contain remote Vars? That would
>> subsume the now() case as well as plenty of others.

> This means we'll translate something like
> explain select * from t where d > now() - '1 day'::interval;
> to
> select * from t where d > $1;

Right.

> The question is how will we reliably determine its typmod (given that I
> read in exprTypmod() comment
> "returns the type-specific modifier of the expression's result type, if
> it can be determined. In many cases, it can't".

I don't think we need to. If exprTypmod() says the typmod is -1,
then that's what it is.

> What do we save if we don't ship whole expression as param, but only
> stable functions? Allowing them seems to be more straightforward.

How so? Right off the bat, you get rid of the need for your own
version of contain_mutable_function. ISTM this approach can probably
be implemented in a patch that's noticeably smaller than what you
have now. It'll likely be touching entirely different places in
postgres_fdw, of course.

regards, tom lane