Alvaro Herrera wrote:
> A trivial, stupid implementation is perhaps not too difficult. The
> problem is getting the smarts right, i.e. an optimized version. You
> certainly don't want to be executing a query against a large table for
> every INSERT on another one, for example; it's better if if you can
> figure out when to skip it.
If you do it using the constraint trigger mechanism, it would work very
similar to foreign keys, I'd imagine. Instead of running a query
against the primary key table, you'd run the check constraint expression.