From: | Jay Levitt <jay(dot)levitt(at)gmail(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | Ants Aasma <ants(at)cybertec(dot)at>, PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>, Francois Deliege <fdeliege(at)gmail(dot)com> |
Subject: | Re: [PATCH] Lazy hashaggregate when no aggregation is needed |
Date: | 2012-03-31 01:56:36 |
Message-ID: | 4F766454.9040909@gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Tom Lane wrote:
> Ants Aasma<ants(at)cybertec(dot)at> writes:
>> A user complained on pgsql-performance that SELECT col FROM table
>> GROUP BY col LIMIT 2; performs a full table scan. ISTM that it's safe
>> to return tuples from hash-aggregate as they are found when no
>> aggregate functions are in use. Attached is a first shot at that.
>
> As I commented in the other thread, the user would be a lot better off
> if he'd had an index on the column in question. I'm not sure it's worth
> complicating the hashagg logic when an indexscan + groupagg would
> address the case better.
Would this patch help in the case where "table" is actually a set-returning
function, and thus can't have an index? (I don't yet know enough about the
tree to know when hashaggs get used). I'm wondering if this is a useful
exception to the "restrictions can't get pushed down through GROUP BYs" rule.
Jay
From | Date | Subject | |
---|---|---|---|
Next Message | Jeff Janes | 2012-03-31 02:07:59 | Re: tracking context switches with perf record |
Previous Message | Tom Lane | 2012-03-30 22:35:30 | Re: Speed dblink using alternate libpq tuple storage |