From: | anthony(dot)shipman(at)symstream(dot)com |
---|---|
To: | pgsql-performance(at)postgresql(dot)org |
Cc: | Claudio Freire <klaussfreire(at)gmail(dot)com> |
Subject: | Re: strange query plan with LIMIT |
Date: | 2011-06-08 08:34:07 |
Message-ID: | 201106081834.07263.anthony.shipman@symstream.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On Wednesday 08 June 2011 17:39, Claudio Freire wrote:
> Of course optimally executing a plan with limit is a lot different
> than one without.
I imagined that limit just cuts out a slice of the query results.
If it can find 80000 rows in 0.5 seconds then I would have thought that
returning just the first 100 of them should be just as easy.
>
> Just... why are you sorting by diag_id?
>
> I believe you would be better off sorting by timestamp than diag_id,
> but I don't know what the query is supposed to do.
The timestamp is only almost monotonic. I need to scan the table in slices and
I use limit and offset to select the slice.
I've forced the query order with some pgsql like:
declare
query character varying;
rec record;
begin
-- PG 8.3 doesn't have the 'using' syntax nor 'return query execute'
execute 'create temporary table tt on commit drop as ' ||
'select diag_id from tdiag ' || v_where;
query = 'select * from tdiag where diag_id in (select * from tt) ' ||
'order by diag_id ' || v_limit || ' ' || v_offset;
for rec in execute query loop
return next rec;
end loop;
end;
--
Anthony Shipman | Life is the interval
Anthony(dot)Shipman(at)symstream(dot)com | between pay days.
From | Date | Subject | |
---|---|---|---|
Next Message | Pavel Stehule | 2011-06-08 08:39:48 | Re: strange query plan with LIMIT |
Previous Message | tv | 2011-06-08 08:33:15 | Re: strange query plan with LIMIT |