Re: In-order pg_dump (or in-order COPY TO)

On 8/27/25 09:10, Dimitrios Apostolou wrote:
>
> On Wednesday 2025-08-27 17:25, Adrian Klaver wrote:

>>
>> For completeness and just in case they may affect the output what do
>> the patches do?
>
> Two patches for speeding up scanning an archive without TOC, like the
> one I'm having (because it is piped into borg, instead of written to
> file). These were activated, but shouldn't matter. They only build the
> TOC in pg_restore's memory.

Are you sure about that?

I just did:

pg_dump -Fc --compress=none --no-toast-compression -d test -U postgres |
borg create --stats --stdin-name pg_file --stdin-user aklaver
--stdin-group aklaver borg_test/::PgTest -

Then:

borg mount borg_test/ mnt_tmp/
cd mnt_tmp/PgTest/

and then:

pg_restore -l pg_file

and I got a TOC.

Or are you streaming the data out of the Borg archive?

>
> https://commitfest.postgresql.org/patch/5809/
> https://commitfest.postgresql.org/patch/5817/
>
> And two patches for speeding up pg_restore like mentioned above, under
> specific arguments that I didn't provide. (one speedup needs --clean,
> and the other needs --freeze).
>
> https://commitfest.postgresql.org/patch/5821/
> https://commitfest.postgresql.org/patch/5826/
>
> IIRC I did not activate them (via --clean) because TRUNCATE fails when
> foreign keys exist. See the discussion threads.
>
>
> Dimitris

--
Adrian Klaver
adrian(dot)klaver(at)aklaver(dot)com