On Fri, Jul 8, 2022 at 11:07 AM Yugo NAGATA <nagata@sraoss.co.jp> wrote:
> On Fri, 08 Jul 2022 09:44:10 +0900 (JST)
> Kyotaro Horiguchi <horikyota.ntt@gmail.com> wrote:
> > At Fri, 8 Jul 2022 01:06:18 +0900, Fujii Masao <masao.fujii@oss.nttdata.com> wrote in
> > > On 2022/07/08 0:33, Tom Lane wrote:
> > > >> On 2022/06/30 10:48, Yugo NAGATA wrote:
> > > >>> When a foreign table has handler but doesn't support TRUNCATE,
> > > >>> an error "cannot truncate foreign table xxx" occurs. So, what
> > > >>> about adding a test this message output? We can add this test
> > > >>> for file_fdw because it is one of the such foreign data wrappers.
> > > > This seems like a fairly pointless expenditure of test cycles
> > > > to me. Perhaps more importantly, what will you do when
> > > > somebody adds truncate support to that FDW?
> > As Fujii-san mentioned below, file_fdw has tests for INSERT/UPDATE and
> > DELETE. If somebody added DELETE to file_fdw, the test for DELETE
> > rejection would be turned into a normal test of the DELETE function.
> > I don't see a difference between TRUNCATE and other updating commands
> > from this point of view.
I agree on this point.
> > > One idea is to create dummy FDW (like foreign_data.sql regression test
> > > does) not supporting TRUNCATE and use it for the test.
> If we want to test foreign table modifications for the FDW framework,
> we will have to add such tests in foreign_data.sql, because foreign
> table modifications are tested only for postgres_fdw and file_fdw.
Rather than doing so, I'd vote for adding a test case to file_fdw, as
proposed in the patch, because that would be much simpler and much
less expensive.
Best regards,
Etsuro Fujita