Michael Paquier <michael.paquier@gmail.com> writes:
> Attached is an updated patch which passes
> check-world in my environment.
Pushed with trivial cosmetic changes to the code, and slightly more
extensive work on the regression test cases.
It strikes me that there might be an argument for returning NULL
rather than raising an error for to_regrole('foo.bar') and
to_regnamespace('foo.bar'). On the other hand, input-syntax issues
result in errors for the other to_reg* functions, so it might
be fine as-is. In any case, relaxing the error rather than adding
one is much easier to justify as a future change, so I left it
as Jim/Michael had it for the moment.
regards, tom lane