From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andrew Gierth <andrew@tao11.riddles.org.uk>
cc: pgsql-bugs@lists.postgresql.org, phb07@apra.asso.fr
Subject: Re: BUG #15129: Problem with UNION/UNION ALL type setting when
 several NULL values before defining the proper type
In-reply-to: <878tago8m2.fsf@news-spur.riddles.org.uk>
References: <152196085023.32649.9916472370480121694@wrigleys.postgresql.org>
 <878tago8m2.fsf@news-spur.riddles.org.uk>
Comments: In-reply-to Andrew Gierth <andrew@tao11.riddles.org.uk>
	message dated "Sun, 25 Mar 2018 09:50:40 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <31858.1521992600.1@sss.pgh.pa.us>
Date: Sun, 25 Mar 2018 11:43:20 -0400
Message-ID: <31859.1521992600@sss.pgh.pa.us>
Precedence: bulk

Andrew Gierth <andrew@tao11.riddles.org.uk> writes:
> "PG" == PG Bug reporting form <noreply@postgresql.org> writes:
>  PG> While migrating a view from another RDBMS,I reached something that
>  PG> looks lie a bug in postgres.

> It's not a bug in the code, though perhaps you could point out a place
> where the documentation could be improved?

It strikes me that section 10.5 doesn't say explicitly that multiple
UNIONs are resolved pairwise.  Someone who expected "x union y union z"
to be resolved holistically, like a 3-way CASE would be, is not going
to be enlightened by that section.  Perhaps an additional example
using this exact situation would be helpful.

> (The SQL standard is of no particular help here since it does not allow
> NULL to appear "bare", except in contextually typed expressions.)

I think the SQL spec does offer considerable support for our pairwise
resolution approach --- nowhere does it suggest that "x union y union z"
is a construct that should be considered as a whole rather than as a
nest of two independent union operations.

			regards, tom lane