From: Tom Lane <tgl@sss.pgh.pa.us>
To: =?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de>
cc: Pg Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: relkind as an enum
In-reply-to: <202602020019.6bseljwhr2k2@alvherre.pgsql>
References: <202602020019.6bseljwhr2k2@alvherre.pgsql>
Comments: In-reply-to =?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de>
	message dated "Mon, 02 Feb 2026 01:42:48 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <401758.1770000067.1@sss.pgh.pa.us>
Date: Sun, 01 Feb 2026 21:41:07 -0500
Message-ID: <401759.1770000067@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/401759.1770000067%40sss.pgh.pa.us>
Precedence: bulk

=?utf-8?Q?=C3=81lvaro?= Herrera <alvherre@kurilemu.de> writes:
> There have been mentions of turning Form_pg_class->relkind into an enum,
> so that we can have compilers provide some more help with
> switch(relkind) blocks.  Here's a quick experiment with that.
> ...
> 0003 is the backend-side change.  This looks generally reasonable,
> though I'm annoyed that I couldn't find a way to coerce the compiler
> into telling me if I had missed some spot.

Yeah.  On the whole I'm dubious that this is worth the code churn,
because I doubt that it will move the needle at all on getting
better compiler warnings.  One reason why I think that is that
I don't think it's okay to assume that what comes off of disk
for the contents of pg_class.relkind is necessarily one of the
valid values: catalog corruption would break that assumption.
So even in places where it's reasonable to list every RELKIND_
value in a switch (and in many of them it looks like that would
be tedium), we'd still need to cater for a default: clause.
And then we'd get no warnings, so what have we gained?

			regards, tom lane