Re: Teaching pg_dump to use NOT VALID constraints

Peter Eisentraut wrote:
> On 11/13/14 5:07 AM, Simon Riggs wrote:
> > On 13 November 2014 00:20, Jim Nasby <Jim(dot)Nasby(at)bluetreble(dot)com> wrote:
> >
> >> Isn't the real use-case here that if constraints were valid when you dumped
> >> then we shouldn't have to *any* re-validate when we load? (Though, we'd have
> >> to be careful of that with CHECK because that can call user code...)
> >
> > Yes, that is the use case this patch would improve considerably.
>
> But your patch doesn't really address that. It just leaves the
> constraints as invalid, and someone will have to revalidate them later
> (how?). What Jim was describing was a mode that creates the constraints
> as valid but doesn't actually validate them. I can see both sides of
> that kind of feature.

This might lead to users introducing invalid data by way of declaring
constants as valid but not checked by the system; if they turn out to be
invalid, the final state is a mess. I would only buy such a feature if
we had some way to pass down the knowledge of the constraint being valid
in the original system through some other means; say emit a CRC of the
copy data in the pg_dump output that can be verified while loading, and
only allow unvalidated constraints to be marked VALID if the sum
matches.

--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services