> One problem I see is that pg_constraint entries can *only* be associated
> with relations; so the table has no way to represent constraints
> associated with domains --- not to mention assertions, which aren't
It's ugly, but one could make the relid 0, and add a typeid which is
non-zero to represent a constraint against a domain. Relation
constraints have typeid 0 and relid as a normal number.
Obviously I prefer unique constraint names mostly for my users. For
some reason they tend to try to make assumptions about a constraint
given the name and have been fooled about what the constraint actually
is more than once due to 'having seen it before elsewhere'.
Is applying a lock on the pg_constraint table really that bad during
creation? Sure, you could only make one constraint at a time -- but
thats the same with relations, types, and a fair number of other things
that are usually created at the same time (or same transaction) as most
constraints will be.