Re: tuple radix sort

> I also thought we should change this cast:
>
>         if (ssup->comparator == ssup_datum_signed_cmp)
>         {
> -               norm_datum1 = orig + ((uint64) PG_INT64_MAX) + 1;
> +               norm_datum1 = orig + (Int64GetDatum(PG_INT64_MAX)) + 1;
>         }
>
> Upthread you mention something else about treating Datum as an
> integer, but I'm not sure if this is what you meant since you didn't
> say. If you have concrete suggestions, feel free to share them.

I think we can keep it as is now. The idea of don't treat datum as an integer comes from:

https://www.postgresql.org/message-id/flat/8246d7ff-f4b7-4363-913e-827dadfeb145%40eisentraut.org

--
Regards,
ChangAo Chen