Re: Sub-optimal plan chosen

On Thu, Sep 10, 2009 at 8:43 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> bricklen <bricklen(at)gmail(dot)com> writes:
> > Is there any other data I can provide to shed some light on this?
>
> The table and index definitions?
>
> The straight indexscan would probably win if the index column order
> were ofid, date instead of date, ofid. I can't tell if you have
> any other queries for which the existing column order is preferable,
> though.
>
> regards, tom lane
>

Changing the order of the WHERE predicates didn't help. The indexes are
mostly defined as single-column indexes, with the exception of the
"statsdaily_unique_idx" one:

statsdaily_id_pk PRIMARY KEY, btree (id)
statsdaily_unique_idx UNIQUE, btree (date, idaf, idsite, ofid, idcreative,
idoptimizer)
statsdaily_date_idx btree (date)
statsdaily_ofid_idx btree (ofid)
statsdaily_ofid_sept2009_idx btree (ofid) WHERE date >= '2009-09-01'::date