Martijn van Oosterhout <kleptog@xxxxxxxxx> writes: > On Tue, May 08, 2007 at 02:14:54PM +0200, Felix Kater wrote: >> There is *no complete* substitute for foreign keys by using *indexes* >> since I'd loose the referencial integrity (whereas for unique contraints >> there *is* a full replacement using indexes)? > A unique index is not a "substitute" for a unique constraint, they're > exactly the same thing. If you drop your constraint and create a unique > index, you're back where you started. You neither added nor removed > anything. Well, actually you added or removed a pg_constraint entry associated with the index ... but either way it's the unique index that really does the work of enforcing uniqueness. regards, tom lane
