On 25.03.24 17:17, Tom Lane wrote:
> PG Bug reporting form <noreply@postgresql.org> writes:
>> I created a foreign table with a serial column and then moved the table to a
>> different schema. The sequence bound to the table column wasn't moved to the
>> new schema. I expected the sequence to be moved as well just like it works
>> for regular tables.
>
> Hmm ... that does seem fairly inconsistent. It looks like the cause
> is in AlterTableNamespaceInternal:
>
> /* Fix other dependent stuff */
> if (rel->rd_rel->relkind == RELKIND_RELATION ||
> rel->rd_rel->relkind == RELKIND_MATVIEW ||
> rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE)
> {
> AlterIndexNamespaces(classRel, rel, oldNspOid, nspOid, objsMoved);
> AlterSeqNamespaces(classRel, rel, oldNspOid, nspOid,
> objsMoved, AccessExclusiveLock);
> AlterConstraintNamespaces(RelationGetRelid(rel), oldNspOid, nspOid,
> false, objsMoved);
> }
>
> It's probably reasonable to have this relkind gating for
> AlterIndexNamespaces, although I'm not sure it saves much to
> skip that, either. But we can have sequences attached to
> foreign tables, and I think maybe constraints as well.
> Maybe it'd be best to just drop this relkind check altogether?
Yes, seems better to remove the relkind check.