remove WITHOUT OIDS syntax for v19

Junwang Zhao <zhjwpku@gmail.com> writes:

WITH OIDS was removed in v12, I'm wondering if we could remove the
WITHOUT OIDS support for v19.

Why?

regards, tom lane

On Tue, Jul 15, 2025 at 11:06:11PM +0800, Junwang Zhao wrote:

WITH OIDS was removed in v12, I'm wondering if we could remove the
WITHOUT OIDS support for v19.

AFAICT this would produce less helpful error messages and might even break
applications, which I don't think is worth it to save ~90 lines.

--
nathan

On Tue, Jul 15, 2025 at 11:11 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:

Junwang Zhao <zhjwpku@gmail.com> writes:

WITH OIDS was removed in v12, I'm wondering if we could remove the
WITHOUT OIDS support for v19.

Why?

Sorry I didn't mention the reason why I think WITHOUT OIDS should be removed.

WITHOUT OIDS has been a backward-compatible syntax for 6 years, so I
think maybe not too many users use it nowadays. Besides, there are
some hints in the code base that `WITHOUT OIDS` and `with (oids =
false)` will be removed someday(see the following diff), so I wonder
which day should it be?

-/* WITHOUT OIDS is legacy only */
OptWith:
WITH reloptions { $$ = $2; }
- | WITHOUT OIDS { $$ = NIL; }
| /*EMPTY*/

- /*
- * This is not a great place for this test,
but there's no other
- * convenient place to filter the option out.
As WITH (oids =
- * false) will be removed someday, this seems
like an acceptable
- * amount of ugly.
- */

regards, tom lane

--
Regards
Junwang Zhao

On Wed, Jul 16, 2025 at 12:49 AM Nathan Bossart
<nathandbossart@gmail.com> wrote:

On Tue, Jul 15, 2025 at 11:06:11PM +0800, Junwang Zhao wrote:

WITH OIDS was removed in v12, I'm wondering if we could remove the
WITHOUT OIDS support for v19.

AFAICT this would produce less helpful error messages and might even break
applications, which I don't think is worth it to save ~90 lines.

Yeah, I agree this might break some legacy applications. I don't know
if we have some strategy/policy for removing such legacy syntax, but
do we want to keep this forever?

--
nathan

--
Regards
Junwang Zhao

On Wed, Jul 16, 2025 at 08:06:44AM +0800, Junwang Zhao wrote:

Sorry I didn't mention the reason why I think WITHOUT OIDS should be removed.

WITHOUT OIDS has been a backward-compatible syntax for 6 years, so I
think maybe not too many users use it nowadays. Besides, there are
some hints in the code base that `WITHOUT OIDS` and `with (oids =
false)` will be removed someday(see the following diff), so I wonder
which day should it be?

This is the zone where we would break stuff only for the sake of
breaking it, so it would be annoying for any existing applications
after a major upgrade. Keeping this code is not really a maintenance
burden AFAIK, we can just let it be.

When the WITH OIDS grammar was removed, there were arguments about
code simplifications and ease of support for table access methods,
which had and still have long-term benefits. We still have
default_with_oids in guc_tables.c, for example. That's a nobrainer to
keep it in the GUC tables, and we avoid breaking the world.
--
Michael