Re: Distinguish publication exclusions in object addresses

On Thu, Sep 17, 2026 at 12:22 PM Peter Smith <smithpb2250(at)gmail(dot)com> wrote:
>
> On Thu, Sep 17, 2026 at 4:11 PM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> >
> > On Thu, Sep 17, 2026 at 12:17 AM Manuel Reyes Bravo
> > <manuelreyesbravo(at)gmail(dot)com> wrote:
> > >
> > > vignesh C <vignesh21(at)gmail(dot)com> wrote:
> > > > Please have a look at the discussion regarding the commit in [1]
> > > > thread. If you still feel that it needs further discussion, we can
> > > > continue discussing your suggestion in the thread at [2], as it is not
> > > > directly related to the topic of this current thread.
> > >
> > > Thanks, I had not seen [2]. Peter raised the nested quotes there on
> > > May 11 with the same kind of example, and you considered them
> > > acceptable, citing REFRESH MATERIALIZED VIEW CONCURRENTLY. So 0002 is
> > > not a new finding, and I withdraw it.
> > >
> > > I also need to correct a number I gave. I wrote that 55 errmsg() calls
> > > in src/backend use "\"%s.%s\"" against 3 that put a quoted qualified
> > > name inside quotes. My search missed calls whose arguments are on
> > > later lines. Reading each errmsg/errdetail/errhint call in full on
> > > REL_19_STABLE, it is 70 against 2: matview.c (the case you cited) and
> > > the EXCEPT message in pg_publication.c.
> > >
> >
> > Seeing the other usage, I am thinking to fix this case for except
> > tables unless some major change is required. We can consider changing
> > REFRESH MATERIALIZED VIEW CONCURRENTLY separately. I have few points
> > for the that patch:
> >
> > * The existing function
> > RelationGetQualifiedRelationName()->get_qualified_objname() used
> > get_namespace_name_or_temp() where as patch used get_namespace_name().
> > That could create the problem with temp tables as follows:
> >
> > Before patch:
> > postgres=# create publication mypub for all tables except(table mytemp);
> > ERROR: cannot specify relation "pg_temp.mytemp" in the publication
> > EXCEPT clause
> > DETAIL: This operation is not supported for temporary tables.
> >
> > After patch:
> > postgres=# create publication mypub for all tables except(table mytemp);
> > ERROR: cannot specify relation "pg_temp_0.mytemp" in the publication
> > EXCEPT clause
> > DETAIL: This operation is not supported for temporary tables.
> >
> > You can notice that temp schema name has _0 appended to it after
> > patch, it seems to append backend's slot number. We can use
> > get_namespace_name_or_temp() instead of get_namespace_name().
> >
> > * After this RelationGetQualifiedRelationName() is dead and is
> > introduced by the commit a49b9cfd72d89a8fd68c90e84c38aa7e2ada756b
> > which introduced this quoting problem. I feel it is better to remove
> > it.
> >
> > Attached fixed both the problems.
> >
>
> Hi Amit.
>
> + relname = psprintf("%s.%s",
> + get_namespace_name_or_temp(RelationGetNamespace(targetrel)),
> + RelationGetRelationName(targetrel));
>
> I had been working off-list on a similar patch to this, so I like the
> direction where this is heading.
>
> OTOH please do not remove the RelationGetQualifiedRelationName function.
>
> Instead, keep the function but replace the body with your
> psprint/get_namespace_name_or_temp logic. There are many other places
> where the errmsg can benefit from having a common function to output
> fully qualified names, but we don't want them all needing to code
> their own psprintf/get_namespace_name_or_temp.
>

Even if we go in that direction, it should be for PG20. In PG19, we
can't extend its usage. So, I prefer to drop it and if we can come up
with a use case for it that is acceptable then we can consider it
again.

--
With Regards,
Amit Kapila.