Re: Changed SRF in targetlist handling

On Wed, May 25, 2016 at 3:55 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Andres Freund <andres@anarazel.de> writes:
>> On 2016-05-25 15:20:03 -0400, Tom Lane wrote:
>>> We could certainly make a variant behavior in nodeFunctionscan.c that
>>> emulates that, if we feel that being exactly bug-compatible on the point
>>> is actually what we want.  I'm dubious about that though, not least
>>> because I don't think *anyone* actually believes that that behavior isn't
>>> broken.  Did you read my upthread message suggesting assorted compromise
>>> choices?
>
>> You mean https://www.postgresql.org/message-id/21076.1464034513@sss.pgh.pa.us ?
>> If so, yes.
>
>> If we go with rewriting this into LATERAL, I'd vote for 2.5 (trailed by
>> option 1), that'd keep most of the functionality, and would break
>> visibly rather than invisibly in the cases where not.
>
> 2.5 would be okay with me.
>
>> I guess you're not planning to work on this?
>
> Well, not right now, as it's clearly too late for 9.6.  I might hack on
> it later if nobody beats me to it.

Curious if this approach will also rewrite:
select generate_series(1,generate_series(1,3)) s;

...into
select s from generate_series(1,3) x cross join lateral generate_series(1,x) s;

another interesting case today is:
create sequence s;
select generate_series(1,nextval('s')), generate_series(1,nextval('s'));

this statement never terminates.  a lateral rewrite of this query
would always terminate with much better defined and well understood
behaviors -- this is good.

merlin