Robert Haas <robertmhaas@gmail.com> writes:
> Looks good, committed with a bit of further cleanup.
I had not actually paid attention to the non-regclass parts of this, and
now that I look, I've got to say that it seems borderline insane to have
chosen to implement regproc/regoper rather than regprocedure/regoperator.
The types implemented here are incapable of dealing with overloaded names,
which --- particularly in the operator case --- makes them close to
useless. I don't think this code was ready to commit.
regards, tom lane