On Mon, 21 Aug 2006, Tom Lane wrote:
> Stephan Szabo <sszabo@megazone.bigpanda.com> writes:
> > Yeah, I think there are a few possibilities around truncate inside a
> > savepoint that's rolledback that we have to be careful of.
>
> Yuck :-(
>
> > If we could mark the entries in some way so we knew whether or not they
> > were made obsolete by a truncate of our own tranasaction or a committed or
> > rolled back past subtransaction of ours, we could probably make both of
> > these work nicely.
>
> That seems much more trouble than it's worth, unless someone can
> convince me that this isn't a corner case with little real-world value.
>
> Furthermore, this still doesn't address the worry about whether there
> are cases where dropping the trigger calls would be inappropriate.
I don't believe there are for foreign keys on the referring side since the
post-truncate case is trivially satisfying the constraint, but I can
imagine that there might exist other uses for deferred triggers for which
one might care.
> I propose just having TRUNCATE check for pending triggers on the
> target tables, and throw an error if there are any.
That sounds reasonable to me, although I don't much use truncate in the
first place.