Re: IMMUTABLE and PARALLEL SAFE function markings

On 2018-11-26 19:13:17 -0500, Stephen Frost wrote:
> Greetings,
>
> * Andres Freund (andres(at)anarazel(dot)de) wrote:
> > On 2018-11-26 19:05:02 -0500, Stephen Frost wrote:
> > > Agreed, but I could see us having a regression test which complains if
> > > it finds any which are marked as immutable but aren't parallel safe.
> >
> > That doesn't help if a user writes a query to review the not parallel
> > safe functions in their installation.
>
> I'm really not sure what you're getting at here..?

You replied to me saying it'd be a bad idea to infer parallel safety
from immutability:

> > > Surely a simple rule could be made that provolatile='i' trumps
> > > proparallel. No need to make them agree.
> > > [...]
> >
> > I think it'd be entirely unacceptable that
> > SELECT * FROM pg_proc WHERE proparallel = 's'
> > wouldn't actually return all the functions that are parallel safe.
>
> Agreed, but I could see us having a regression test which complains if
> it finds any which are marked as immutable but aren't parallel safe.

I'm saying that that doesn't address my concern.

> Parallel safe functions should be marked as such. Immutable functions
> should be marked as such. We should not assume that one implies the
> other, nor should we operate as if they do.
>
> My suggestion for a regression test was to make PG developers really
> think about if their new immutable functions should also be marked as
> parallel safe, in the event that they forget to mark it as such. If
> it's really immutable and not parallel safe, then they need to adjust
> the expected regression test output (and we can all see it...).

See http://archives.postgresql.org/message-id/20181126234521.rh3grz7aavx2ubjv%40alap3.anarazel.de

Greetings,

Andres Freund