Thanks for your answer !
Yes, I set up enable_seqscan = off !
Thanks !
On 7 Jul 2003 at 8:46, Stephan Szabo wrote:
>
> On Mon, 7 Jul 2003 [EMAIL PROTECTED] wrote:
>
> > Hi,
> >
> > when I do an explain on a certain query, I have this answer :
> >
> >
> > QUERY PLAN
> > --------------------------------------------------------------------------------------
> > Aggregate (cost=100017927.48..100017927.48 rows=1 width=8)
> > -> Seq Scan on stats_daily_2003
> > (cost=100000000.00..100017927.47 rows=1 width=8)
> > Filter: ((id_compte = 29075) AND (periode = '07-07-2003'::date))
> > (3 rows)
> >
> >
> > Well, I don"t really undestand the meaning of
> >
> > cost=100000000.00..100017927.47
> >
> > and
> >
> > cost=100017927.48..100017927.48
> >
> > I guess this result is relatively bad.
>
> This probably actually implies that you've done something like set
> enable_seqscan=off at some point which makes the costs of seqscans very high
> to discourage them. In any case, this plan is still expected to be much
> more expensive than the below even when not taking that into account. You
> may also want to do explain analyze to get real world numbers rather than
> just the estimates.
>
> > The same query on a similar table but with an index give :
> > QUERY PLAN
> > --------------------------------------------------------------------------------------------
> > ----------------
> > Aggregate (cost=9.60..9.60 rows=1 width=8)
> > -> Index Scan using ap_stats_daily_2003_save on
> > stats_daily_2003_save (cost=0.00..9.59 rows=2 width=8)
> > Index Cond: ((periode = '07-07-2003'::date) AND (id_compte =
> > 29075))
> > (3 rows)
> >
> > Could you confirm me the second explain is "better" than the first one
> > ?
>
>
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster
Jean-Arthur Silve
EuroVox
4, Place Félix Eboué
75583 Paris Cedex 12
T : +33 1 44670505
F : +33 1 44670519
---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to [EMAIL PROTECTED] so that your
message can get through to the mailing list cleanly
