On Thu, Jan 23, 2014 at 3:46 PM, Bill Moran <wmoran@xxxxxxxxxxxxxxxxx> wrote: > It was suggested that > leaving the constraints as NOT VALID might affect the planner, causing > it to use less optimal plans because it doesn't think it can trust > the constraint. Is this true? AFAICT the planner doesn't currently rely on FOREIGN KEY constriants for anything, so there's no downside to leaving those NOT VALID. UNIQUE constraints affect the planner the most, but they must always be valid anyway. If you use table inheritance (partitioning), then valid CHECK constraints are necessary to use that effectively. > It has also been suggested that manually changing the status to valid > in the catalog without going through the validation process could cause > problems It's unsupported: if you break something when manually messing with the system catalog, you get to keep the pieces and people will just tell you "we told you so". But if you know what you're doing, it's OK. Just make sure you double-check the source code that you're not missing something critical that it does. Regards, Marti -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
