Re: general purpose array_sort

On Fri, Oct 25, 2024 at 8:02 PM Junwang Zhao <zhjwpku(at)gmail(dot)com> wrote:
>
> On Fri, Oct 25, 2024 at 1:19 AM Aleksander Alekseev
> <aleksander(at)timescale(dot)com> wrote:
> >
> > Hi,
> >
> > > I can accept this outcome though an optional three-valued boolean sort order (ascending and descending only) I'd argue is worth keeping. null value placement too I guess, three-valued boolean (nulls_first).
> >
> > Perhaps these optional arguments deserve separate discussions. I
> > suggest merging something everyone agrees on first. This will simplify
> > the review process and allow us to deliver value to the users quickly.
> > Arguments like `reverse => true` and `nulls_first => true` can always
> > be implemented and added as separate patches.
>
> As this patch uses the tuplesort infrastructure, we need to supply the
> sortOperator, sortCollation and nullsFirstFlag, I tend to agree with
> David. I admit that the parsing part is not good, so I will remove it
> by using two boolean parameters Jian suggested earlier.
>
> Will send out another version by tomorrow.

Based on the previous discussion, I split it into two patches in V8.

0001 is the general sort part without `is_ascending` or `nulls_first`,
the sort order is determined by the "<" operator of the element type.
It also cached the type entry of both eletyp and the corresponding
array type.

0002 adds the `is_ascending` and `nulls_first` part, it now uses
two boolean parameters instead of parsing one text parameter.

>
> >
> > --
> > Best regards,
> > Aleksander Alekseev
>
>
>
> --
> Regards
> Junwang Zhao

--
Regards
Junwang Zhao