Re: Implement targetlist SRFs using ROWS FROM() (was Changed SRF in targetlist handling)

Andres Freund <andres@anarazel.de> writes:
> On 2017-01-18 16:56:46 -0500, Tom Lane wrote:
>> Andres Freund <andres@anarazel.de> writes:
>> I have not actually looked at 0003 at all yet.  So yeah, please post
>> for review after you're done rebasing.

> Here's a rebased and lightly massaged version.

I've read through this and made some minor improvements, mostly additional
comment cleanup.  One thing I wanted to ask about:

@@ -4303,7 +4303,7 @@ inline_function(Oid funcid, Oid result_type, Oid result_collid,
 
     /*
      * Forget it if the function is not SQL-language or has other showstopper
-     * properties.  (The nargs check is just paranoia.)
+     * properties.  (The nargs and retset checks are just paranoia.)
      */
     if (funcform->prolang != SQLlanguageId ||
         funcform->prosecdef ||

I thought this change was simply wrong, and removed it; AFAIK it's
perfectly possible to get here for set-returning functions, since
the planner does expression simplification long before it worries
about splitting out SRFs.  Did you have a reason to think differently?

Other than that possible point, I think the attached is committable.

			regards, tom lane