On Thu, May 22, 2008 at 9:38 PM, Gurjeet Singh <singh.gurjeet@gmail.com> wrote:
> I have seen discussions in the past on the -hackers mailing list about
> deprecating the usage of SRFs in select list, but honestly, I don't see this
> usage being deprecated anytime soon; it's pretty useful, and I would call it
> an extension to the SQL language. And as long as it's nit buggy, and ALA it
> doesn't surprise people in a wrong way, I think such usage will remain.
I think it's useful too...I have a good example of how it can simplify
a query in the archives.
My main gripe with 'SRF in select' is multiple SRF in the select list...
select a(), b() from foo;
The current behavior gives you LCM(a(), b()) * foo records...I think
the expected behavior would be a * b * foo (foo being number of
records in foo).
merlin