Re: WITH ORDINALITY versus column definition lists

On 20 November 2013 22:46, Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk> wrote:
>>>>>> "Tom" == Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
>
> Tom> 1. Reinsert HEAD's prohibition against directly combining WITH
> Tom> ORDINALITY with a coldeflist (with a better error message and a
> Tom> HINT suggesting that you can get what you want via the TABLE
> Tom> syntax).
>
> That gets my vote.
>

Yeah that seems preferable to option #2, which just seems to open up a
whole can of worms.

However, I think I would quickly find it a PITA that it kept telling
me to wrap it in a TABLE() construct. It would seem like the "TABLE"
was just an unnecessary noise word (from a user perspective). Could we
simply support an alias list after the ORDINALITY, in addition to the
coldeflist? For example:

select * from array_to_set(array['one', 'two']) as (f1 int,f2 text)
with ordinality as t(a1,a2,a3);

That could be regarded as an implicit TABLE() construct, but the
syntax would be closer to the non-coldeflist case:

[ LATERAL ] function_name ( [ argument [, ...] ] )
[ WITH ORDINALITY ] [ [ AS ] alias [ ( column_alias [, ...] ) ] ]

[ LATERAL ] function_name ( [ argument [, ...] ] )
[ AS ] [ alias ] ( column_definition [, ...] )
[ WITH ORDINALITY ] [ [ AS ] alias [ ( column_alias [, ...] ) ] ]

Regards,
Dean