John Naylor <jcnaylor(at)gmail(dot)com> writes:
> It seems most of the time the FooRelationId labels are predictable,
> although not as pristine as the Anum_* constants. One possibility that
> came to mind is to treat these like pg_type OID #defines -- have a
> simple rule that can be overridden for historical reasons.
Meh. I'd just as soon avoid having some catalogs done one way and
some another.
regards, tom lane