To: Joe Conway <mail@joeconway.com>
Cc: Markus Bertheau <twanger@bluetwanger.de>,
	pgsql-sql@postgresql.org
Subject: Re: ARRAY() returning NULL instead of ARRAY[] resp. {} 
In-reply-to: <42923762.7010708@joeconway.com> 
References: <1116873219.7306.0.camel@localhost.localdomain>
	<42923762.7010708@joeconway.com>
Comments: In-reply-to Joe Conway <mail@joeconway.com>
	message dated "Mon, 23 May 2005 13:04:50 -0700"
Date: Tue, 24 May 2005 00:06:00 -0400
Message-ID: <16774.1116907560@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Joe Conway <mail@joeconway.com> writes:
> Markus Bertheau wrote:
>> why does SELECT ARRAY(SELECT 1 WHERE FALSE) return NULL instead of
>> ARRAY[] resp. '{}'?

> Why would you expect an empty array instead of a NULL?

I think he's got a good point, actually.  We document the ARRAY-with-
parens-around-a-SELECT syntax as

	The resulting one-dimensional array will have an element for
	each row in the subquery result, with an element type matching
	that of the subquery's output column.

To me, that implies that a subquery result of no rows generates a
one-dimensional array of no elements, not a null array.

This is not the same as

SELECT ARRAY[(SELECT 1 WHERE FALSE)];

We define a scalar subquery that returns no rows as returning null, so
this is equivalent to

SELECT ARRAY[NULL];

which *ought* to yield an array containing a single NULL element,
but since we can't yet handle arrays containing nulls we punt and
return a null array value.  That's wrong too ... but it's a different
issue.  The point Markus is complaining about seems like it should
be easily fixable.

			regards, tom lane