Re: BUFFER_LOCK_EXCLUSIVE is used in ginbuildempty().

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Peter Geoghegan <pg(at)heroku(dot)com>
Cc: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, Kyotaro HORIGUCHI <horiguchi(dot)kyotaro(at)lab(dot)ntt(dot)co(dot)jp>, Pg Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: BUFFER_LOCK_EXCLUSIVE is used in ginbuildempty().
Date: 2014-07-17 19:54:31
Message-ID: 10710.1405626871@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Peter Geoghegan <pg(at)heroku(dot)com> writes:
> On Thu, Jul 17, 2014 at 7:47 AM, Alvaro Herrera
> <alvherre(at)2ndquadrant(dot)com> wrote:
>> I don't understand the point of having these GIN_EXCLUSIVE / GIN_SHARED
>> symbols. It's not like we could do anything different than
>> BUFFER_LOCK_EXCLUSIVE etc instead. It there was a GinLockBuffer() it
>> might make more sense to have specialized symbols, but as it is it seems
>> pointless.

> It's a pattern common to the index AMs. I think it's kind of pointless
> myself, but as long as we're doing it we might as well be consistent.

I think that to the extent that these symbols are used in APIs above the
direct buffer-access layer, they are useful --- for example using
BT_READ/BT_WRITE in _bt_search calls seems like a useful increment of
readability. GIN seems to have less of that than some of the other AMs,
but I do see GIN_SHARE being used that way in some calls.

BTW, there's one direct usage of BUFFER_LOCK_EXCLUSIVE in the GIST code
as well, which should probably be replaced with GIST_EXCLUSIVE if we're
trying to be consistent.

regards, tom lane

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Fabien COELHO 2014-07-17 20:13:24 Re: gaussian distribution pgbench
Previous Message Tom Lane 2014-07-17 19:38:06 Re: [v9.5] Custom Plan API