On Tue, 6 Jan 2026 at 19:22, Aleksander Alekseev
<aleksander@tigerdata.com> wrote:
>
> Hi,
>
> > > Kirill, are you working on this patch?
> >
> > Yes, PFA.
> >
> > > I've not studied in detail,
> > > but looks like it would require making
> > > char_decrement()/char_increment() and btcharskipsupport() all use
> > > CharGetDatum()/DatumGetChar() and switch to using SCHAR_MIN/SCHAR_MAX
> > > rather than 0/UCHAR_MAX.
> >
> > This is also a possible change, and can be a separate patch. I will
> > try to also work on this this week.
>
> v1 looks incomplete to me. As I understand, Tom proposed to get rid of
> UInt8 conversions as well. Are you going to implement this idea as
> well?
>
Hmm, v1 looks good and self-contained to me. Like, anyway, making two
commits (one for signed Int8 and one for unsigned) here is better for
sake of atomicy?
Anyway, I can see there are users of UInt8GetDatum, which are [0] and
forks of Greenplum. So, I am not super-sure removing UInt8* is
desirable.
[0] https://github.com/duckdb/pg_duckdb/blob/main/src/pgduckdb_types.cpp#L230
--
Best regards,
Kirill Reshke
