From: Tom Lane <tgl@sss.pgh.pa.us>
To: Tomas Vondra <tomas.vondra@2ndquadrant.com>
cc: PostgreSQL-development <pgsql-hackers@postgresql.org>
Subject: Re: [HACKERS] user-defined numeric data types triggering ERROR:
 unsupported type
In-reply-to: <b5cede7b-354b-83e6-ab1f-7a1bf8a232fa@2ndquadrant.com>
References: <b68441b6-d18f-13ab-b43b-9a72188a4e02@2ndquadrant.com>
 <9389.1506003842@sss.pgh.pa.us>
 <4b444362-4240-49e0-3523-724e5932b060@2ndquadrant.com>
 <24767.1506180446@sss.pgh.pa.us>
 <d526eb53-dbd6-d414-8e5d-21e644ce4549@2ndquadrant.com>
 <14597.1520038563@sss.pgh.pa.us>
 <42f4e37e-f923-6ea9-4172-cdec5a8c2d25@2ndquadrant.com>
 <b5cede7b-354b-83e6-ab1f-7a1bf8a232fa@2ndquadrant.com>
Comments: In-reply-to Tomas Vondra <tomas.vondra@2ndquadrant.com>
	message dated "Sat, 03 Mar 2018 19:03:07 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <11915.1520127473.1@sss.pgh.pa.us>
Date: Sat, 03 Mar 2018 20:37:53 -0500
Message-ID: <11916.1520127473@sss.pgh.pa.us>
Precedence: bulk

Tomas Vondra <tomas.vondra@2ndquadrant.com> writes:
> Here is v2 of the fix. It does handle all the convert_to_scalar calls
> for various data types, just like the numeric one did in v1 with the
> exception of bytea.

Pushed with some adjustments.

> The bytea case is fixed by checking that the boundary values are
> varlenas. This seems better than checking for BYTEAOID explicitly, which
> would fail for custom varlena-based types. At first I've been thinking
> there might be issues when the data types has mismatching ordering, but
> I don't think the patch makes it any worse.

I didn't like this one bit.  It's unlike all the other cases, which accept
only specific type OIDs, and there's no good reason to assume that a
bytea-vs-something-else comparison operator would have bytea-like
semantics.  So I think it's better to punt, pending the invention of an
API to let the operator supply its own convert_to_scalar logic.

> I've also added a bunch of regression tests, checking each case. The
> bytea test it should cause segfault on master, of course.

I was kind of underwhelmed with these test cases, too, so I didn't
commit them.  But they were good for proving that the bytea bug
wasn't hypothetical :-)

			regards, tom lane