Re: counting algorithm for incremental matview maintenance

From: Kevin Grittner <kgrittn(at)ymail(dot)com>
To: Merlin Moncure <mmoncure(at)gmail(dot)com>
Cc: "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: counting algorithm for incremental matview maintenance
Date: 2013-05-15 18:18:35
Message-ID: 1368641915.94135.YahooMailNeo@web162903.mail.bf1.yahoo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Merlin Moncure <mmoncure(at)gmail(dot)com> wrote:
> Kevin Grittner <kgrittn(at)ymail(dot)com> wrote:
>> Merlin Moncure <mmoncure(at)gmail(dot)com> wrote:
>>
>>> #1 issue I have with current matview functionality is locking.
>>> currently refresh takes out an access exclusive lock.  so,
>>> question is, do you think your proposal will be such that it
>>> will no longer require taking out full lock for refresh
>>> purposes (either incremental or otherwise)?
>>
>> The right thread for *that* question is "Differential
>> (transactional) REFRESH"; however, I might as well say here that
>> I don't think we want to get rid of the (faster) version that
>> just replaces the current heap when we add the (slower) option
>> to REFRESH it transactionally.
>
> sorry, didn't notice that thread.  agreed, that seems good
> candidate for user input to refresh command to manage the
> tradeoff.
>
> well, do you expect the application of differential refresh to be
> automatic?

I expect considerable bikeshedding on this, but my preference would
be to allow syntax for specifying which technique is desired on
the REFRESH command, and in the absence of specification I would
prefer that it default to the current technique for a matview which
is not yet populated and the transactional technique for a
populated matview.  We could associate some property with the
matview for default REFRESH technique, but I don't know whether
that's worth the trouble.

> lockless + differential refresh would be game changer in terms of
> how I build up data for analytics.

Yeah, it's definitely something I would have liked to have in the
initial release, but I tried to keep the scope very limited given
how little time there was left in the release cycle when I got a
chance to start on it.  Adding this seemed to be just a little too
much.

--
Kevin Grittner
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Peter Geoghegan 2013-05-15 18:32:29 Re: Parallel Sort
Previous Message Noah Misch 2013-05-15 18:12:22 Re: Parallel Sort