On Tue, Jul 21, 2015 at 11:29 AM, Paul Ramsey <pramsey@cleverelephant.ca> wrote:
> On July 21, 2015 at 11:22:12 AM, Tom Lane (tgl@sss.pgh.pa.us) wrote:
>
> No, *not* populated first-time-through, because that won't handle any of
> the CREATE, DROP, or UPGRADE cases. It's also doing a lot of work you
> might never need. I was thinking of "populate on demand", that is, first
> time you need to know whether function X is shippable, you find that out
> and then cache the answer (whether it be positive or negative).
>
>
> Roger that. Off to the races..
Attached, reworked with a local cache. I felt a little dirty sticking
the cache entry right in postgres_fdw.c, so I broke out all my
nonsense into shippable.c.
Thanks!
P