Re: [bug?] Missed parallel safety checks, and wrong parallel safety

On Fri, Jul 23, 2021 at 6:55 PM Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>
> On Wed, Jul 21, 2021 at 11:55 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> > I see here we have a mix of opinions from various people. Dilip seems
> > to be favoring the approach where we provide some option to the user
> > for partitioned tables and automatic behavior for non-partitioned
> > tables but he also seems to have mild concerns about this behavior.
> > OTOH, Greg and Hou-San seem to favor an approach where we can provide
> > an option to the user for both partitioned and non-partitioned tables.
> > I am also in favor of providing an option to the user for the sake of
> > consistency in behavior and not trying to introduce a special kind of
> > invalidation as it doesn't serve the purpose for partitioned tables.
> > Robert seems to be in favor of automatic behavior but it is not very
> > clear to me if he is fine with dealing differently for partitioned and
> > non-partitioned relations. Robert, can you please provide your opinion
> > on what do you think is the best way to move forward here?
>
> I thought we had agreed on handling partitioned and unpartitioned
> tables differently, but maybe I misunderstood the discussion.
>

I think for the consistency argument how about allowing users to
specify a parallel-safety option for both partitioned and
non-partitioned relations but for non-partitioned relations if users
didn't specify, it would be computed automatically? If the user has
specified parallel-safety option for non-partitioned relation then we
would consider that instead of computing the value by ourselves.

Another reason for hesitation to do automatically for non-partitioned
relations was the new invalidation which will invalidate the cached
parallel-safety for all relations in relcache for a particular
database. As mentioned by Hou-San [1], it seems we need to do this
whenever any function's parallel-safety is changed. OTOH, changing
parallel-safety for a function is probably not that often to matter in
practice which is why I think you seem to be fine with this idea. So,
I think, on that premise, it is okay to go ahead with different
handling for partitioned and non-partitioned relations here.

[1] - https://www.postgresql.org/message-id/OS0PR01MB5716EC1D07ACCA24373C2557941B9%40OS0PR01MB5716.jpnprd01.prod.outlook.com

--
With Regards,
Amit Kapila.