> Thanks for testing! You're right, looks like in the current implementation in
> case of backwards scan there is one unnecessary extra step forward. It seems
> this mistake was made, since I was concentrating only on the backward scans
> with a cursor, and used not exactly correct approach to wrap up after a scan
> was finished. Give me a moment, I'll tighten it up.
Thanks. Looking forward to it. I think I found some other strange behavior. Given the same table as in my previous e-mail, the following queries also return inconsistent results. I spent some time trying to debug it, but can't easily pinpoint the cause. It looks like it also skips over one value too much, my guess is during _bt_skippage call in _bt_skip?
Perhaps a question: when stepping through code in GDB, is there an easy way to pretty print for example the contents on an IndexTuple? I saw there's some tools out there that can pretty print plans, but viewing tuples is more complicated I guess.
-- this one is OK
postgres=# select distinct on (a) a,b from a where b>1;
a | b
---+---
1 | 2
2 | 2
3 | 2
4 | 2
5 | 2
(5 rows)
-- this one is not OK, it seems to skip too much
postgres=# select distinct on (a) a,b from a where b=2;
a | b
---+---
1 | 2
3 | 2
5 | 2
(3 rows)