On 3/2/2024 02:06, Alena Rybakina wrote:
> On 01.02.2024 08:00, jian he wrote:
> I added your code to the patch.
Thanks Alena and Jian for the detailed scrutiny!
A couple of questions:
1. As I see, transformAExprIn uses the same logic as we invented but
allows composite and domain types. Could you add a comment explaining
why we forbid row types in general, in contrast to the transformAExprIn
routine?
2. Could you provide the tests to check issues covered by the recent (in
v.15) changes?
Patch 0001-* in the attachment incorporates changes induced by Jian's
notes from [1].
Patch 0002-* contains a transformation of the SAOP clause, which allows
the optimizer to utilize partial indexes if they cover all values in
this array. Also, it is an answer to Alexander's note [2] on performance
degradation. This first version may be a bit raw, but I need your
opinion: Does it resolve the issue?
Skimming through the thread, I see that, in general, all issues have
been covered for now. Only Robert's note on a lack of documentation is
still needs to be resolved.
[1]
https://www.postgresql.org/message-id/CACJufxGXhJ823cdAdp2Ho7qC-HZ3_-dtdj-myaAi_u9RQLn45g%40mail.gmail.com
[2]
https://www.postgresql.org/message-id/CAPpHfduJtO0s9E%3DSHUTzrCD88BH0eik0UNog1_q3XBF2wLmH6g%40mail.gmail.com
--
regards,
Andrei Lepikhov
Postgres Professional