Pavel Stehule <pavel.stehule@gmail.com> writes:
> this is just small note about length of this patch. This patch was
> significantly smaller then he solved problem with derivate types for
> compound types - it should to solve problem described in this thread
> http://stackoverflow.com/questions/7634704/declare-variable-of-composite-type-in-postgresql-using-type
Well, I think what that example shows is that there's a good reason for
plpgsql_parse_wordtype and plpgsql_parse_cwordtype to handle the
PLPGSQL_NSTYPE_ROW case, which they could do based on the tdtypeid from
the row's tupdesc. Still isn't going to run to anything like 500 lines
of new code, nor justify a grammar rewrite that risks introducing new
bugs. The existing code doesn't need to special-case type names that
are also plpgsql keywords, and I'd just as soon not introduce an
assumption that there's no overlap there.
regards, tom lane