Matthias van de Meent <boekewurm+postgres@gmail.com> writes:
> On Mon, 23 Oct 2023, 19:34 Tom Lane, <tgl@sss.pgh.pa.us> wrote:
>> After ruminating on this for awhile, here's a straw-man proposal:
>> ...
> How does this work w.r.t. concurrently created tables that contain the
> domain?
It wouldn't change that at all I think. I had noticed that we'd
probably need to tweak validateDomainConstraint() to ensure it applies
the same semantics that INSERT/UPDATE do --- although with Isaac's
idea to enable better tracking of which constraints will fail on NULL,
maybe just a blind application of the constraint expression will still
be close enough.
I agree that concurrent transactions can create violations of the new
constraint, but (a) that's true now, (b) I have no good ideas about
how to improve it, and (c) it seems like an independent problem.
regards, tom lane