Re: UNION and pg_restore

Thanks, Tom, for your helpful explanation and suggestions.

> Now you're doubtless wondering why Postgres doesn't dodge this ambiguity
> for you.

This is exactly what I was wondering, of course. And I follow the reasoning behind why it cannot, at present. If Postgres can't ensure that the view definition is valid SQL, though, what about the (seemingly more manageable) idea of providing some kind of notice when that definition is not re-loadable? Perhaps pg_dump could do this?

Maybe that sounds like hand-holding, and I suspect the response will be "always test your backup before you need it!" And certainly I've learned my lesson about that. Believe me, I don't mean to "request" anything here, merely raise the point that for users that rely heavily on interdependent VIEWs, a cascade of errors flowing back from pg_restore can make for a pretty frightening moment.

Thanks,
Bryan