Andres Freund <andres@anarazel.de> writes:
> On 2020-10-20 14:16:12 -0400, Tom Lane wrote:
>> I'd make that point against the whole proposal. There's nothing here that
>> can't be done with current_setting() + set_config().
> The one case where I can see SET support being useful even without
> config support is to allow for things like
> ALTER DATABASE somedatabase SET search_path += 'myapp';
Hmm, yeah, that's fractionally less easy to build from spare parts
than the plain SET case.
But I think there are more definitional hazards than you are letting
on. If there's no existing pg_db_role_setting entry, what value
exactly are we += 'ing onto, and why?
regards, tom lane
