Re: inherit support for foreign tables

> > I wonder what shall be the cases when foreign table is on a server which
> does not support *all* SQL features.
> >
> > Does a FDW need to have the possible inherit options mentioned in its
> documentation for this patch?
>
> The answer is no, in my understanding.  The altering operation simply
> declares some chages for foreign tables in the inheritance tree and does
> nothing to the underlying storages.
>
It seems to me Atri mention about the idea to enforce constraint when
partitioned foreign table was referenced...

BTW, isn't it feasible to consign FDW drivers its decision?
If a foreign table has a check constraint (X BETWEEN 101 AND 200),
postgres_fdw will be capable to run this check on the remote server,
however, file_fdw will not be capable. It is usual job of them when
qualifiers are attached on Path node.
Probably, things to do is simple. If the backend appends the configured
check constraint as if a user-given qualifier, FDW driver will handle it
appropriately.

Thanks,
--
NEC OSS Promotion Center / PG-Strom Project
KaiGai Kohei <kaigai@ak.jp.nec.com>


> -----Original Message-----
> From: pgsql-hackers-owner@postgresql.org
> [mailto:pgsql-hackers-owner@postgresql.org] On Behalf Of Etsuro Fujita
> Sent: Tuesday, January 28, 2014 1:08 PM
> To: Atri Sharma; David Fetter
> Cc: pgsql-hackers@postgresql.org
> Subject: Re: [HACKERS] inherit support for foreign tables
>
> (2014/01/28 0:55), Atri Sharma wrote:
> >> On 27-Jan-2014, at 21:03, David Fetter <david@fetter.org> wrote:
> >>> On Mon, Jan 27, 2014 at 05:06:19PM +0900, Etsuro Fujita wrote:
> >>> Hi Hanada-san,
> >>>
> >>> While still reviwing this patch, I feel this patch has given enough
> >>> consideration to interactions with other commands, but found the
> >>> following incorrect? behabior:
> >>>
> >>> postgres=# CREATE TABLE product (id INTEGER, description TEXT);
> >>> CREATE TABLE postgres=# CREATE FOREIGN TABLE product1 () INHERITS
> >>> (product) SERVER fs OPTIONS (filename '/home/foo/product1.csv',
> >>> format 'csv'); CREATE FOREIGN TABLE postgres=# ALTER TABLE product
> >>> ALTER COLUMN description SET STORAGE EXTERNAL;
> >>> ERROR:  "product1" is not a table or materialized view
> >>>
> >>> ISTN the ALTER TABLE simple recursion mechanism (ie
> >>> ATSimpleRecursion()) should be modified for the ALTER COLUMN SET
> >>> STORAGE case.
> >>
> >> This points to a larger discussion about what precisely foreign
> >> tables can and cannot inherit from local ones.  I don't think that a
> >> generic solution will be satisfactory, as the PostgreSQL FDW could,
> >> at least in principle, support many more than the CSV FDW, as shown above.
> >>
> >> In my estimation, the outcome of discussion above is not a blocker
> >> for this
>
> I just thought that among the structures that local tables can alter, the
> ones that foreign tables also can by ALTER FOREIGN TABLE are inherited,
> and the others are not inherited.  So for the case as shown above, I thought
> that we silently ignore executing the ALTER COLUMN SET STORAGE command for
> the foreign table.  I wonder that would be the first step.
>
> > I wonder what shall be the cases when foreign table is on a server which
> does not support *all* SQL features.
> >
> > Does a FDW need to have the possible inherit options mentioned in its
> documentation for this patch?
>
> The answer is no, in my understanding.  The altering operation simply
> declares some chages for foreign tables in the inheritance tree and does
> nothing to the underlying storages.
>
> Thanks,
>
> Best regards,
> Etsuro Fujita
>
>
> --
> Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make
> changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-hackers