[HACKERS] fix possible optimizations in ATExecAttachPartition() - Mailing list pgsql-hackers

From Jeevan Ladhe
Subject [HACKERS] fix possible optimizations in ATExecAttachPartition()
Date
Msg-id CAOgcT0Neue5EgZFHz9qLox1kNENFOv90XqhXzNm97+VPba9Ucw@mail.gmail.com
Whole thread Raw
Responses Re: [HACKERS] fix possible optimizations in ATExecAttachPartition()  (Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>)
List pgsql-hackers
Hi,

I was doing some testing for my default partitioning work, and I realized that
there seem to be a some optimization possible (and I think we should really have
this optimization) for logic around handling the "key IS NOT NULL" constraints
followed by predicate_implied_by() call.

1. Consider, if predicate_implied_by() returns false, in that case skip_validate
is left set to false, and we really do not want to do lots of stuff followed
this decision just to prove that the scan cannot be avoided(which is already
decided in this case) if there is no NOT NULL constraint on the partition key
column.

2. Further, if we have already decided not to skip the scan, we should really have
a break in following if block:

if (!partition_accepts_null &&
(partattno == 0 ||
!bms_is_member(partattno, not_null_attrs)))
skip_validate = false;


I have made above changes in the attached patch.
PFA, and let me know if I am missing something here.

Regards,
Jeevan Ladhe

Attachment

pgsql-hackers by date:

Previous
From: Petr Jelinek
Date:
Subject: Re: [HACKERS] Refreshing subscription relation state inside atransaction block
Next
From: Amit Langote
Date:
Subject: Re: [HACKERS] fix possible optimizations in ATExecAttachPartition()