On Tue, Jul 23, 2013 at 9:38 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> If it weren't that we've been speculating for years about deprecating
> SRFs-in-tlists once we had LATERAL, I would personally consider this
> patch DOA in this form. If we do think we'll probably deprecate that
> feature, then not extending WITH ORDINALITY to such cases is at least
> defensible. On the other hand, considering that we've yet to ship a
> release supporting LATERAL, it's probably premature to commit to such
> deprecation --- we don't really know whether people will find LATERAL
> to be a convenient and performant substitute.
I guess I'd sort of assumed that the plan was to continue accepting
SRFs in tlists but rewrite them as lateral joins, rather than getting
rid of them altogether. IIUC that would simplify some things inside
the executor. I'd be a bit more reluctant to just ban SRFs in target
lists outright.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company