Re: set_config() documentation clarification - Mailing list pgsql-hackers

From Tom Lane
Subject Re: set_config() documentation clarification
Date
Msg-id 3425846.1609883280@sss.pgh.pa.us
Whole thread Raw
In response to Re: set_config() documentation clarification  (Bruce Momjian <bruce@momjian.us>)
List pgsql-hackers
Bruce Momjian <bruce@momjian.us> writes:
> On Tue, Jan  5, 2021 at 08:23:15PM +0100, Joel Jacobson wrote:
>> út 5. 1. 2021 v 18:51 odesílatel Bruce Momjian <bruce@momjian.us> napsal:
>>> I think I tried to make this feature more visible a few years ago and
>>> some people said we might remove it someday, so don't do that.  If that
>>> is false, I think we can make it more prominent.

>> I think it's false.

> Oh, I know it is useful.  My pg_cryptokey uses it:
>     https://momjian.us/download/pgcryptokey/

People use it because we don't have a better substitute.  But I think
the way forward is to create a better substitute, not to encourage
abuse of custom GUCs ... and this is an abuse.  Two good reasons why
are

* there's no way to declare a data type for such a variable from SQL

* guc.c is not designed to scale to large numbers of variables,
nor to heavy update activity.

I've not looked at Pavel's "schema variables" patch lately, but
maybe people who are interested in this should be pushing that
forward.

            regards, tom lane



pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: plpgsql variable assignment with union is broken
Next
From: Peter Geoghegan
Date:
Subject: Re: Deleting older versions in unique indexes to avoid page splits