Обсуждение: BUG #18541: Reattaching a partition referenced by a foreign key fails with an error

The following bug has been logged on the website:

Bug reference:      18541
Logged by:          Alexander Lakhin
Email address:      exclusion@gmail.com
PostgreSQL version: 17beta2
Operating system:   Ubuntu 22.04
Description:

The following script:
CREATE TABLE t1 (a int, PRIMARY KEY (a));
CREATE TABLE t (a int, PRIMARY KEY (a), FOREIGN KEY (a) REFERENCES t1)
PARTITION BY LIST (a);
ALTER TABLE t ATTACH PARTITION t1 FOR VALUES IN (1);
ALTER TABLE t DETACH PARTITION t1;
ALTER TABLE t ATTACH PARTITION t1 FOR VALUES IN (1);

ends up with the error complaining about check triggers:
ERROR:  XX000: could not find ON INSERT check triggers of foreign key
constraint 16400
LOCATION:  GetForeignKeyCheckTriggers, tablecmds.c:11260

Reproduced on REL_15_STABLE (starting from f4566345c) .. master.

On 2024-Jul-15, Tender Wang wrote:

> PG Bug reporting form <noreply@postgresql.org> 于2024年7月15日周一 21:02写道：
> 
> > The following script:
> > CREATE TABLE t1 (a int, PRIMARY KEY (a));
> > CREATE TABLE t (a int, PRIMARY KEY (a), FOREIGN KEY (a) REFERENCES t1)
> > PARTITION BY LIST (a);
> > ALTER TABLE t ATTACH PARTITION t1 FOR VALUES IN (1);
> > ALTER TABLE t DETACH PARTITION t1;
> > ALTER TABLE t ATTACH PARTITION t1 FOR VALUES IN (1);
> >
> > ends up with the error complaining about check triggers:
> > ERROR:  XX000: could not find ON INSERT check triggers of foreign key
> > constraint 16400
> > LOCATION:  GetForeignKeyCheckTriggers, tablecmds.c:11260

> I saw the same error in [1]. I guess it is same issue.
> I send a patch in [1], but it may need a more work.

I think this schema is nuts.  Do you have a practical use for something
like this?

I am tempted to fix this by dictating that you can't join a table as a
partition if the partitioned table contains an FK that references that
table.

FWIW the patch I have fixes all the other reported problems with FKs and
partitioned tables, but not this one.

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/

On 2024-Aug-07, Alvaro Herrera wrote:

> I think this schema is nuts.  Do you have a practical use for something
> like this?

FWIW if you try to set this up the other way around, you get a different
error:

create table t (a int primary key) partition by list (a);
create table t1 partition of t for values in (1);
alter table t add foreign key (a) references t1;
ERROR:  cannot ALTER TABLE "t1" because it is being used by active queries in this session

so I'm not bothered by the restriction in functionality.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/

On 2024-Jul-15, Tender Wang wrote:

> PG Bug reporting form <noreply@postgresql.org> 于2024年7月15日周一 21:02写道：

> > The following script:
> > CREATE TABLE t1 (a int, PRIMARY KEY (a));
> > CREATE TABLE t (a int, PRIMARY KEY (a), FOREIGN KEY (a) REFERENCES t1)
> > PARTITION BY LIST (a);
> > ALTER TABLE t ATTACH PARTITION t1 FOR VALUES IN (1);

I propose to reject this with the attached patch, which I intend to
backpatch all the way down to 12.

FWIW, it's not the same problem that Jehan-Guillaume described in [1],
even though the error message being thrown is the same.

[1] https://postgr.es/m/20230705233028.2f554f73@karst

-- 
Álvaro Herrera               48°01'N 7°57'E  —  https://www.EnterpriseDB.com/

On 2024-Aug-08, Tender Wang wrote:

> Thanks for explanation. The attached patch looks good to me.

Thanks for looking!  I have pushed this now.

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/