> -----Original Message-----
> From: pgsql-hackers-owner@postgresql.org [mailto:pgsql-hackers-
> owner@postgresql.org] On Behalf Of Andrew Dunstan
> Sent: Monday, April 18, 2011 1:43 PM
> To: Tom Lane
> Cc: Peter Eisentraut; pgsql-hackers
> Subject: Re: [HACKERS] ORDER BY 1 COLLATE
>
> On 04/18/2011 04:20 PM, Tom Lane wrote:
> > Peter Eisentraut<peter_e@gmx.net> writes:
> >> This came from a review by Noah Misch a great while ago:
> >> test=> SELECT b FROM foo ORDER BY 1 COLLATE "C";
> >> ERROR: 42804: collations are not supported by type integer
> >> According to SQL92, this should be supported. Do we want to bother?
> It
> >> doesn't look hard to fix, so it's really only a question of whether
> this
> >> would be useful, or its absence would be too confusing.
> > The ORDER BY 1 business seems to me to be legacy anyway. I'm not
> > inclined to put in even more hacks to make strange combinations work
> > there --- I think we're likely to find ourselves painted into a
> corner
> > someday as it is.
> >
> >
>
> It's likely to be used by SQL generators if nothing else, and I've been
> known to use it as a very convenient shorthand. It would seem to me
> like
> quite a strange inconsistency to allow order by n with some qualifiers
> but not others.
I use order by <result_set_column_number> a lot, especially when result_set_column is a complicated expression.