BUG #3543: ARRAY(SELECT ...) contruct yields NULL without rows

On Thu, Aug 16, 2007 at 11:11:49AM +0000, Jens Schicke wrote:

The following bug has been logged online:

Bug reference: 3543
Logged by: Jens Schicke
Email address: j.schicke@asco.de
PostgreSQL version: 8.2.4
Operating system: GNU/Linux
Description: ARRAY(SELECT ...) contruct yields NULL without rows
Details:

SELECT ARRAY(SELECT 1 WHERE 1 = 0) IS NULL; -- true

And just what would you have it return? An array with no elements is by
definition NULL.
--
Decibel!, aka Jim Nasby decibel@decibel.org
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)

Decibel! <decibel@decibel.org> writes:

On Thu, Aug 16, 2007 at 11:11:49AM +0000, Jens Schicke wrote:

SELECT ARRAY(SELECT 1 WHERE 1 = 0) IS NULL; -- true

And just what would you have it return? An array with no elements is by
definition NULL.

No, there is a difference:

regression=# select null::int4[];
int4
------

(1 row)

regression=# select '{}'::int4[];
int4
------
{}
(1 row)

I'm inclined to think that Jens is right and it'd be more consistent to
return the latter when the SELECT returns no rows. It'd be a behavioral
change more than a bug fix, though, so I would not suggest back-patching
it.

[ looks at code... ] It looks like the implementation problem is for
ExecScanSubPlan to know what the array element type is supposed to be.
Normally it finds that out from the sub-select's returned rows; but
if there aren't any then it'll have to do something else.

regards, tom lane