Re: Restrict publishing of partitioned table with a foreign table as partition

On Mon, 7 Apr 2025 at 09:43, Sergey Tatarintsev
<s(dot)tatarintsev(at)postgrespro(dot)ru> wrote:
>
> 07.04.2025 03:27, Álvaro Herrera пишет:
>
> On 2025-Apr-01, Shlok Kyal wrote:
>
> I have modified the comment in create_publication.sgml and also added
> comment in the restrictions section of logical-replication.sgml.
> I have also added a more detailed explanation in comment of
> 'check_foreign_tables'
>
> I have attached the updated v11 patch.
>
> Sadly I don't have time to describe the changes proposed here right now,
> but I'll do that early tomorrow. (Some minor changes are still needed,
> particularly the comments to publication_check_foreign_parts which are
> mostly unchanged from what your patch has. I'll do another review round
> tomorrow.)
>
> Hello!
>
> I looked at the latest patch again and found one more place for list_free(). Also look at additional test case:
>
> diff --git a/src/backend/catalog/pg_publication.c b/src/backend/catalog/pg_publication.c index 51e463c112b..7fcc191feb9 100644 --- a/src/backend/catalog/pg_publication.c +++ b/src/backend/catalog/pg_publication.c @@ -442,6 +442,7 @@ GetPubPartitionOptionRelations(List *result, PublicationPartOpt pub_partopt, result = lappend_oid(result, partOid); } + list_free(all_parts); } else result = lappend_oid(result, relid); diff --git a/src/test/regress/sql/publication.sql b/src/test/regress/sql/publication.sql index 49c9d98b668..e56aebc397a 100644 --- a/src/test/regress/sql/publication.sql +++ b/src/test/regress/sql/publication.sql @@ -1296,6 +1296,14 @@ SELECT pubname, tablename FROM pg_publication_tables WHERE schemaname in ('sch3' -- foreign partition ALTER PUBLICATION pub1 SET (publish_via_partition_root); +CREATE SCHEMA sch5; +CREATE SCHEMA sch6; +CREATE TABLE sch6.tmain(id int) PARTITION BY RANGE(id); +CREATE TABLE sch5.part1 PARTITION OF sch6.tmain FOR VALUES FROM (0) TO (10) PARTITION BY RANGE(id); +CREATE FOREIGN TABLE sch6.part2 PARTITION OF sch5.part1 FOR VALUES FROM (0) TO (5) SERVER fdw_server; +CREATE PUBLICATION pub4 FOR TABLES IN SCHEMA sch5 WITH (publish_via_partition_root); +SELECT pubname, tablename FROM pg_publication_tables WHERE pubname = 'pub4' ORDER BY pubname, tablename; + DROP PUBLICATION pub1; DROP PUBLICATION pub2; DROP PUBLICATION pub3;
>
> I think this is a wrong assumption:
>
> ScanKeyInit(&key[keycount++], Anum_pg_class_relispartition, BTEqualStrategyNumber, F_BOOLEQ, BoolGetDatum(false));
>
> In this case sch5.part1 is partitioned table, but it also partition of table in different schema
>

Hi Sergey,

Thanks for reviewing the patch. I have addressed the comment in the
latest patch shared in [1].

[1]: https://www.postgresql.org/message-id/CANhcyEW0QMiJXMqpPFRHni-q0Rm4R0hpZ0LdaqA%3DF3wvDUU6sQ%40mail.gmail.com

Thanks and Regards,
Shlok Kyal