Re: ATTACH PARTITION locking documentation for DEFAULT partitions

On Sat, 17 Apr 2021 at 00:03, Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
> PFA an updated patch. I've updated the wording of the previous patch,
> and also updated this section in alter_table.sgml, but with different
> wording, explictly explaining the process used to validate the altered
> default constraint.

I had to squint at this:

+ALTER TABLE measurement_default ADD CONSTRAINT excl_y2008m02
+   CHECK ( (logdate >= DATE '2008-02-01' AND logdate < DATE
'2008-03-01') IS FALSE );

I tried your example and it does not work.

set client_min_messages = 'debug1';
create table rp (dt date not null) partition by range(dt);
create table rp_default partition of rp default;
alter table rp_default add constraint rp_default_chk check ((dt >=
'2022-01-01' and dt < '2023-01-01') is false);
create table rp_2022 partition of rp for values from ('2022-01-01') to
('2023-01-01');

There's no debug message to indicate that the constraint was used.

Let's try again:

alter table rp_default drop constraint rp_default_chk;
drop table rp_2022;
alter table rp_default add constraint rp_default_chk check (not (dt >=
'2022-01-01' and dt < '2023-01-01'));
create table rp_2022 partition of rp for values from ('2022-01-01') to
('2023-01-01');
DEBUG:  updated partition constraint for default partition
"rp_default" is implied by existing constraints

The debug message indicates that it worked as expected that time.

But to be honest, I don't know why you've even added that. There's not
even an example on how to add a DEFAULT partition, so why should we
include an example of how to add a CHECK constraint on one?

I've spent a bit of time hacking at this and I've come up with the
attached patch.

David