Re: Avoiding superfluous buffer locking during nbtree backwards scans

On 2024-11-13 00:55, Peter Geoghegan wrote:
> On Sun, Nov 10, 2024 at 11:36 PM Masahiro Ikeda
> <ikedamsh(at)oss(dot)nttdata(dot)com> wrote:
>> Thanks! The change made it easier for me to understand.
>
> As follow-up to all of the recent work in this area, I'd like to add
> this wrapper function to return the next item from so->currPos.
>
> The wrapper function has extra assertions, compared to what we do
> already. It's slightly more defensive, and IMV slightly clearer.

Thanks, I agree with adding the function for refactoring and including
assertions for moreLeft or moreRight.

One thing I was concerned about is that "if (scan->xs_want_itup)" was
changed to "if (so->currTuples)". However, this isn’t an issue because
so->currTuples is only initialized if scan->xs_want_itup is set to
true in btrescan(), and it improves consistency with other functions
for index-only scans.

I also confirmed that make check-world passes with the patch.

Regards,
--
Masahiro Ikeda
NTT DATA CORPORATION