On Mon, Jan 18, 2021, at 19:33, Tom Lane wrote:
> On second thought, a catalog is overkill; it'd only be useful if the data
> could change after initdb, which this data surely cannot. The right way
> to expose such info to SQL is with a set-returning function reading a
> constant table in the C code, a la pg_get_keywords(). That way takes a
> whole lot less infrastructure and is just as useful to SQL code.
+1