Re: ALTER TABLE .. DETACH PARTITION CONCURRENTLY

On Sat, Apr 10, 2021 at 01:42:26PM -0500, Justin Pryzby wrote:
> On Sun, Mar 21, 2021 at 03:01:15PM -0300, Alvaro Herrera wrote:
> > > But note that it doesn't check if an existing constraint "implies" the new
> > > constraint - maybe it should.
> >
> > Hm, I'm not sure I want to do that, because that means that if I later
> > have to attach the partition again with the same partition bounds, then
> > I might have to incur a scan to recheck the constraint. I think we want
> > to make the new constraint be as tight as possible ...
>
> If it *implies* the partition constraint, then it's at least as tight (and
> maybe tighter), yes ?
>
> I think you're concerned with the case that someone has a partition with
> "tight" bounds like (a>=200 and a<300) and a check constraint that's "less
> tight" like (a>=100 AND a<400). In that case, the loose check constraint
> doesn't imply the tighter partition constraint, so your patch would add a
> non-redundant constraint.
>
> I'm interested in the case that someone has a check constraint that almost but
> not exactly matches the partition constraint, like (a<300 AND a>=200). In that
> case, your patch adds a redundant constraint.
>
> I wrote a patch which seems to effect my preferred behavior - please check.

On Sat, Apr 10, 2021 at 02:13:26PM -0500, Justin Pryzby wrote:
> I suppose the docs should be updated for the exact behavior, maybe even without
> this patch:
>
> |A <literal>CHECK</literal> constraint
> |that duplicates the partition constraint is added to the partition.

I added this as an Opened Item, since it affects user-visible behavior:
whether or not a redundant, non-equal constraint is added.

--
Justin