On Sat, Mar 9, 2019 at 12:03 AM Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
>
> On 2019-Mar-08, Amit Langote wrote:
>
> > On Fri, Mar 8, 2019 at 11:09 PM Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
>
> > > I'm not sure about copying the same to ddl.sgml. Why is that needed?
> > > Update is not DDL.
> >
> > Hmm, maybe because there's already a huge block of text describing
> > certain limitations of UPDATE row movement under concurrency?
>
> Uh, you're right, there is. That seems misplaced :-( I'm not sure it
> even counts as a "limitation"; it seems to belong to the NOTES section
> of UPDATE rather than where it is now.
>
> > Actually, I remember commenting *against* having that text in
> > ddl.sgml, but it got in there anyway.
>
> We can move it now ...
>
> > > ddl.sgml does say this: "Partitions can also be
> > > foreign tables, although they have some limitations that normal tables
> > > do not; see CREATE FOREIGN TABLE for more information." which suggests
> > > that the limitation might need to be added to create_foreign_table.sgml.
> >
> > Actually, that "more information" never got added to
> > create_foreign_table.sgml. There should've been some text about the
> > lack for tuple routing at least in PG 10's docs, but I guess that
> > never happened.
>
> Sigh.
>
> Since version 10 is going to be supported for a few years still, maybe
> we should add it there.
>
> > Should we start now by listing this UPDATE row movement limitation?
>
> I think we should, yes.
Attached find 3 patches -- for PG 10, 11, and HEAD. I also realizes
that a description of PARTITION OF clause was also missing in the
Parameters section of CREATE FOREIGN TABLE, which is fixed too.
Thanks,
Amit