Re: PREPARE/EXECUTE across backends?

Neil Conway <neilc(at)samurai(dot)com> writes:
> The decision to store prepared statements per-backend, rather than in
> shared memory, was made deliberately. In fact, an early version of the
> PREPARE/EXECUTE patch (written by Karel Zak) stored prepared statements
> in shared memory. But I decided to remove this, because:
> [ several good reasons ]

Another issue is that we currently don't have a mechanism for flushing
query plans when they become obsolete (eg, an index is added or
removed). Locally-cached plans are relatively easy to refresh: just
start a fresh session. A shared plan cache would retain bogus plans
forever, short of a postmaster restart.

Obviously we need a mechanism for detecting and handling cached-plan
invalidations, and I hope someone will get around to that soon.
But we *cannot* consider a shared plan cache until that mechanism
exists.

If I recall correctly, Karel's original shared plan cache also triggered
a lot of concern about contention for the shared data structure ...
I'm not convinced that it would be a big bottleneck, but there's
definitely an issue to think about there ...

regards, tom lane