To: Peter Eisentraut <peter_e@gmx.net>
Cc: milimeter@163.com,
	PostgreSQL Development <pgsql-hackers@postgresql.org>
Subject: Re: [GENERAL] Problem about pgsql's column alias 
In-reply-to: <Pine.LNX.4.44.0304141455000.1919-100000@peter.localdomain> 
References: <Pine.LNX.4.44.0304141455000.1919-100000@peter.localdomain>
Comments: In-reply-to Peter Eisentraut <peter_e@gmx.net>
	message dated "Mon, 14 Apr 2003 18:40:01 +0200"
Date: Mon, 14 Apr 2003 13:00:55 -0400
Message-ID: <10849.1050339655@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Peter Eisentraut <peter_e@gmx.net> writes:
> Tom Lane writes:
>> AS is not optional in Postgres, and will never be so unless we choose to
>> drop support for postfix operators, which I consider an unlikely choice.

> Well, we could drop the support for postfix operators in aliasing clauses
> that omit the AS.

I think it would be a bad idea in any case; it would mask errors too
easily.  For example, you meant to say "SELECT a + b, ..." but wrote
"SELECT a b, ...".  Or how about this:
	SELECT 'foo'::character varying
Is "varying" an alias or part of the type name?

We've recently seen at least one comparable case where someone
unintentionally wrote what looked to the parser like a FROM-clause
alias, and was mystified at the behavior (I forget the details, but it
was just a week or two back).  Omitting AS in FROM-clauses is tolerable
because the possible syntax in that area is fairly restricted, but I
think the spec authors were just plain foolish to allow it in the SELECT
targetlist.

			regards, tom lane