Re: unnesting multirange data types

On Tue, Jul 13, 2021 at 5:07 PM Justin Pryzby <pryzby(at)telsasoft(dot)com> wrote:
> On Tue, Jul 13, 2021 at 03:11:16PM +0300, Alexander Korotkov wrote:
> > On Sun, Jul 11, 2021 at 1:20 AM Justin Pryzby <pryzby(at)telsasoft(dot)com> wrote:
> > > On Sun, Jul 11, 2021 at 01:00:27AM +0300, Alexander Korotkov wrote:
> > > > On Sat, Jul 10, 2021 at 7:34 PM Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> wrote:
> > > > > On 2021-Jun-27, Alexander Korotkov wrote:
> > > > >
> > > > > > BTW, I found some small inconsistencies in the declaration of
> > > > > > multirange operators in the system catalog. Nothing critical, but if
> > > > > > we decide to bump catversion in beta3, this patch is also nice to
> > > > > > push.
> > > > >
> > > > > Hmm, I think you should push this and not bump catversion. That way,
> > > > > nobody is forced to initdb if we end up not having a catversion bump for
> > > > > some other reason; but also anybody who initdb's with beta3 or later
> > > > > will get the correct descriptions.
> > > > >
> > > > > If you don't push it, everybody will have the wrong descriptions.
> > > >
> > > > True, but I'm a bit uncomfortable about user instances with different
> > > > catalogs but the same catversions. On the other hand, initdb's with
> > > > beta3 or later will be the vast majority among pg14 instances.
> > > >
> > > > Did we have similar precedents in the past?
> > >
> > > It seems so.
> > >
> > > Note in particular 74ab96a45, which adds a new function with no bump.
> > > Although that one may not be a good precedent to follow, or one that's been
> > > followed recently.
> >
> > Justin, thank you very much for the summary.
> >
> > Given we have similar precedents in the past, I'm going to push the
> > patch [1] to master and pg14 if no objections.
>
> To be clear, do you mean with or without this hunk ?
>
> - oprrest => 'multirangesel', oprjoin => 'scalargtjoinsel' },
> + oprrest => 'multirangesel', oprjoin => 'scalarltjoinsel' },

I mean with this hunk unless I hear objection to it.

The implementations of scalarltjoinsel and scalargtjoinsel are the
same. And I don't think they are going to be changed on pg14.

------
Regards,
Alexander Korotkov