Re: Microoptimization of Bitmapset usage in postgres_fdw - Mailing list pgsql-hackers

From Daniel Gustafsson
Subject Re: Microoptimization of Bitmapset usage in postgres_fdw
Date
Msg-id BB9AAEE6-294F-4296-A8F2-70AEE9EF3E70@yesql.se
Whole thread Raw
In response to Re: Microoptimization of Bitmapset usage in postgres_fdw  (Ashutosh Bapat <ashutosh.bapat@enterprisedb.com>)
Responses Re: Microoptimization of Bitmapset usage in postgres_fdw
List pgsql-hackers
> On 30 May 2018, at 09:36, Ashutosh Bapat <ashutosh.bapat@enterprisedb.com> wrote:
>
> On Tue, May 29, 2018 at 9:10 PM, Daniel Gustafsson <daniel@yesql.se> wrote:
>> There are a couple of places in postgres_fdw where we check if the Bitmapset
>> has multiple members using bms_num_members(), without storing the returned
>> count.  The attached patch instead use bms_membership() which is optimized for
>> just that usecase, and (IMO) makes for clearer code.
>
> +1 for that change. Some of those usages of bms_num_members() were
> added by me. Sorry for that. It was mostly because I wasn't aware of
> bms_membership() when I wrote that code. May be we should add a
> comment in the prologue of bms_num_members() like "Note: if the
> callers is interested only knowing whether the bitmapset has 0, 1 or
> more members, it should call bms_membership().". I understand that
> bms_membership() resides just below bms_num_members(), but 1.
> bms_membership doesn't sound like it would tell me that 2.
> bms_membership's prologue refers to bms_num_members, which should have
> been the other way; we want the developers to use bms_membership
> instead of bms_num_members(), when they land on bms_num_members. It's
> less likely that somebody landing on bms_membership would want to use
> bms_num_members().

That makes sense, I’ve added a second patch to this thread which expands the
comment on bms_num_members to make it clearer.

> I am not sure if this can b e squeezed into HEAD right now. It looks
> safe to do so. But in case not, please add this to the next commitfest
> so that it's not forgotten.

Will do, thanks for reviewing.

cheers ./daniel


Attachment

pgsql-hackers by date:

Previous
From: Antonin Houska
Date:
Subject: Re: Incorrect visibility test function assigned to snapshot
Next
From: Michael Paquier
Date:
Subject: Re: PATCH pass PGOPTIONS to pg_regress