Re: speedup COPY TO for partitioned table.

On Wed, Jan 22, 2025 at 6:54 AM Melih Mutlu <m(dot)melihmutlu(at)gmail(dot)com> wrote:
>
> Hi Jian,
>
> Thanks for the patch.
>
> jian he <jian(dot)universality(at)gmail(dot)com>, 19 Ara 2024 Per, 15:03 tarihinde şunu yazdı:
>>
>> attached copy_par_regress_test.sql is a simple benchmark sql file,
>> a partitioned table with 10 partitions, 2 levels of indirection.
>> The simple benchmark shows around 7.7% improvement in my local environment.
>
>
> I confirm that the patch introduces some improvement in simple cases like the one you shared. I looked around a bit to understand whether there is an obvious reason why copying from a partitioned table is not allowed, but couldn't find one. It seems ok to me.

hi. melih mutlu
thanks for confirmation.

> I realized that while both "COPY <partitioned_table> TO..." and "COPY (SELECT..) TO..." can return the same set of rows, their orders may not be the same. I guess that it's hard to guess in which order find_all_inheritors() would return tables, and that might be something we should be worried about with the patch. What do you think?
>

in the
find_all_inheritors->find_inheritance_children->find_inheritance_children_extended

find_inheritance_children_extended we have
"""
if (numoids > 1)
qsort(oidarr, numoids, sizeof(Oid), oid_cmp);
"""

so the find_all_inheritors output order is deterministic?