Peter Geoghegan <peter@2ndquadrant.com> writes:
> I felt that this was quite unnecessary because of the limited scope of
> the patch, and because this raises thorny issues of both semantics and
> implementation. Tom agreed with this general view - after all, this
> patch exists for the express purpose of having a well-principled way
> of obtaining the various fields across lc_messages settings. So I
> don't see that we have to do anything about making a constraint_schema
> available.
Or in other words, there are two steps here: first, create
infrastructure to expose the fields that we already provide within the
regular message text; then two, consider adding new fields. The first
part of that is a good deal less controversial than the second, so let's
go ahead and get that part committed.
regards, tom lane