Re: UNNEST with multiple args, and TABLE with multiple funcs

On Mon, Dec 2, 2013 at 11:26 PM, Noah Misch <noah(at)leadboat(dot)com> wrote:
> On Mon, Dec 02, 2013 at 08:56:03PM -0500, Tom Lane wrote:
>> Noah Misch <noah(at)leadboat(dot)com> writes:
>> > ... I propose merely changing the syntax to "TABLE FOR ROWS (...)".
>>
>> Ugh :-(. Verbose and not exactly intuitive, I think. I don't like
>> any of the other options you listed much better. Still, the idea of
>> using more than one word might get us out of the bind that a single
>> word would have to be a fully reserved one.
>>
>> > ROWS FROM
>>
>> This one's a little less awful than the rest. What about "ROWS OF"?
>
> I had considered ROWS OF and liked it, but I omitted it from the list on
> account of the shift/reduce conflict from a naturally-written Bison rule.
> Distinguishing it from a list of column aliases takes extra look-ahead. We
> could force that to work. However, if we ever wish to allow an arbitrary
> from_item in the list, it would become ambiguous: is this drawing rows from
> "a" or just using an alias with a column list?
>
> WITH a AS (SELECT oid FROM pg_am ORDER BY 1) SELECT * FROM rows of(a, a);
>
> ROWS FOR is terse and conflict-free. "FOR" evokes the resemblance to looping
> over the parenthesized section with the functions acting as generators.

I like the idea of using ROWS + some additional word. I think I
mildly prefer Tom's suggestion of ROWS FROM to your suggestion of ROWS
FOR, but I can live with either one.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company