At 02:17 PM 25/11/2003 +0200, you wrote:

>OK, so the problem is not in view.
>
>select distinct currency from v_stat_history where sdate between
>'1.10.2003' and '3.10.2003' and accid=2 order by row
>(9 sec)
>select currency from v_stat_history where sdate between '1.10.2003' and
>'3.10.2003' and accid=2 order by row
>(1.5 sec)
>
>The statstics and PLAN are exactly the same. Why so big time difference?

The first one has to eliminate duplicates - that involves many more hits
per row because it first has to find all of the duplicates. Also, if you
have an index on currency and it only has a few possible values, you should
drop the index and possibly recreate it as a composite of currency + the PK.

Particularly if most of the rows have the same currency value, you'll
probably find a horribly long dup chain for that index when you run
gstat. Long dup chain = death to performance.

heLen