Re: PL/PGSQL: Dynamic Record Introspection

Bruce Momjian <pgman@candle.pha.pa.us> writes:
> It seems like a cool feature.

The fundamental problem with it is that plpgsql isn't designed to deal
with dynamically changing data types.  The patch as submitted contained
some hacks that sort of dealt with this in some cases (I don't think it
covered them all), but really some serious thought is going to have to
go into making that work.  It'd be good to do, because the existing
RECORD-variable feature already creates cases where it's needed; but
AFAICS it's not something that can be fixed off-the-cuff.

            regards, tom lane