Re: Non-text mode for pg_dumpall

Thanks Jian.

On Wed, 12 Feb 2025 at 12:45, jian he <jian(dot)universality(at)gmail(dot)com> wrote:
>
> On Wed, Feb 12, 2025 at 1:17 AM Mahendra Singh Thalor
> <mahi6run(at)gmail(dot)com> wrote:
> >
> > >
> > > There are some tests per https://commitfest.postgresql.org/52/5495, I
> > > will check it later.
>
> hi.
> the cfbot failure is related to function _tocEntryRequired
>
> if (strcmp(te->desc, "DATABASE") == 0 ||
> strcmp(te->desc, "DATABASE PROPERTIES") == 0)
> {
> - if (ropt->createDB)
> + if (ropt->createDB || AH->format != archNull)
> return REQ_SCHEMA;
> else
> return 0;
>
> for restoring multiple databases:
> in v16 implementation: pg_restore even if you do not specify --create,
> it actually did what pg_restore --create option does.
>
> if there are multiple databases in the archive:
> to make the pg_restore --file output is usable, the output file need
> have \connect and CREATE DATABASE
> command. that is exactly what --create option would do.
> pg_restore --file behavior need align with pg_restore --dbname.
> therefore pg_restore restoring multiple databases will use --create option.
>
>
> we can either error out (pg_fatal) saying
> restoring multiple databases requires the pg_restore --create option.
> Or we can add a pg_log_info saying
> pg_restore --create option will be set to true while restoring
> multiple databases.

In my earlier version, I was giving an error if --create option was
not specified.

I think it will be good and more preferable if we give an error
without the --create option if dump was taken from pg_dumpall. Even
though there is a single database in the dump of pg_dumpall, it is
possible that a particular database hasn't been created.
Ex: -d postgres and we have db1 dump in file. In this case, we have
only one database dump but this database has not been created.
If the user wants to restore a single database, then the user should
use a single database dump file. Forcefully adding --create option is
not a good idea, instead we will give an error to the user and let him
correct the inputs.

Apart from the above handling, I fixed all the pending review comments
in this patch and made some more changes.

Here, I am attaching an updated patch for review and testing.

--
Thanks and Regards
Mahendra Singh Thalor
EnterpriseDB: http://www.enterprisedb.com