On 2014-01-06 10:29:06 -0500, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > Does anybody have an opinion about introducing generic pseudotype IO
> > functions?
>
> Yes: -1.
Ok, fine with me.
> > Pseudotype.c/pg_proc.h are slowly growing a number of pretty
> > useless/redundant copy&pasted functions... Most for cases that are
> > pretty damn unlikely to be hit by users not knowing what they do.
>
> That's hardly the largest cost associated with inventing a new pseudotype.
> Nor are there lots of new pseudotypes coming down the pike, anyway.
Robert suggested modelling the lookup of changeset extraction output
callbacks after fdw's FdwRoutine, that's why I am wondering about
it. I noticed while reviewing that I so far had borrowed fdw's C
routines which didn't seem like such a nice thing to do...
> > What about adding a pseudotype_in/out that just error out with a generic
> > message?
>
> This will break some of the function sanity checks in opr_sanity,
> I believe. Yeah, we could lobotomize that, but I don't see any benefit.
Yes. But there's precedent in refcursor using text's routines...
(it's in type_sanity, but whatever)
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services