Re: ALTER TABLE .. DETACH PARTITION CONCURRENTLY

On 2021-Mar-21, Justin Pryzby wrote:

> On Fri, Mar 19, 2021 at 10:57:37AM -0300, Alvaro Herrera wrote:
> > > Also, it "fails to avoid" adding duplicate constraints:
> > >
> > > Check constraints:
> > > "c" CHECK (i IS NOT NULL AND i > 1 AND i < 2)
> > > "cc" CHECK (i IS NOT NULL AND i >= 1 AND i < 2)
> > > "p1_check" CHECK (true)
> > > "p1_i_check" CHECK (i IS NOT NULL AND i >= 1 AND i < 2)
> >
> > Do you mean the "cc" and "p1_i_check" one? I mean, if you already have
>
> No, I started with c and cc, and it added the broken constraint p1_check (which
> you say you've fixed) and the redundant constraint p1_i_check. I guess that's
> what you meant.

Yes, that's what I meant.

> > a constraint in the partition that duplicates the partition constraint,
> > then during attach we still create our new constraint? I guess a
> > solution to this would be to scan all constraints and see if any equals
> > the expression that the new one would have. Sounds easy enough now that
> > write it out loud.
>
> But it looks like DetachAddConstraintIfNeeded already intended to do that:
>
> + if (equal(constraintExpr, thisconstr))
> + return;

Hah, so I had already done it, but forgot.

> Actually, it appears your latest notpatch resolves both these issues.

Great.

> 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 ...

--
Álvaro Herrera 39°49'30"S 73°17'W