On Sun, Mar 16, 2025 at 9:50 PM jian he <jian.universality@gmail.com> wrote:
>
> hi.
>
> before commit 4618045bee4a6d3efcb489c319649d8dd9aaa738 ([0])
It seems that patch is treating int2vector and oidvector as scalar
types.
>
> select array_sort(array(select '1 4'::int2vector union all select '1
> 2'::int2vector));
> array_sort
> --------------------------
> [1:2][0:1]={{1,2},{1,4}}
> (1 row)
>
This query:
select array(select '1 4'::int2vector union all select '1 2'::int2vector);
get
[1:2][0:1]={{1,4},{1,2}}
So the behavior of array_sort is consistent.
> after
>
> select array_sort(array(select '1 4'::int2vector union all select '1
> 2'::int2vector));
> array_sort
> ---------------
> {"1 2","1 4"}
> (1 row)
>
> now look closer, the second is the expected result...
> I didn't dig deeper why commit 4618045bee made this patch result correct,
> but I guess it would be best to include such a test case,
> so I've attached a patch.
>
No objection for the test case, thanks.
>
> [0] https://git.postgresql.org/cgit/postgresql.git/commit/?id=4618045bee4a6d3efcb489c319649d8dd9aaa738
--
Regards
Junwang Zhao
