> There is no alternative, unless you want the command to be
> non-roll-back-able.
Well, you can do a cluster-type table duplication...
> > Someone can
> > make it more efficient in regards to constraint checks, etc. in the
> > future if they want -- I don't intend to.
>
> It'd be nice if you at least ensure that all the constraints are checked
> in a single pass over the table (not one per constraint). Right offhand
> I do not see why they couldn't be checked in the same pass that does the
> UPDATE. For extra credit, detect that the default expression is
> immutable or stable, and do the checks *once* not once per row.
And check domain constraints...?
Chris
