"Florian G. Pflug" <fgp@phlo.org> writes:
> Still, doing that SELECT seems rather silly since NEW and OLD already
> contain the required information. So I still believe that having
> something like record_name() and record_types() would be useful. And at
> least these functions have less of an issue with the type system...
Yeah. I don't have any objection in principle to providing such
functions; I'm just wondering how far that really goes towards
solving real-world problems.
regards, tom lane