Fabio,
Sergey,

The test that I showed was in one * prg.
The environment was that which exists when I enter vfp.
The time of the request is of 6 seconds!
Display status To File Test1.txt
ftp://195.146.205.14/Utilisateurs/Bernhart/


I put the same code in my application.
Environment codebook (Ed.leave)
Incredible!The time of the request is 0.49!
Display status To spins Test.txt
ftp://195.146.205.14/Utilisateurs/Bernhart/

I don't understand.
Bell, century, confirm, exclusive are different.
who can explain me?

Bernhart




>Give me this sum and count value of :
>
>

>SELECT COUNT(*),SUM(Ncat) FROM Sum_Categ_per_periode
>

>
>I think join is not optimizable, and query need to scan fixed subset of the full cartesian product of two table .
>
>Only possible ( hard ) solution , but it could not be better, is with manual query, with use of data statistics ( like unique values, etc ).
>This need partial query, expression index and a scan with internal
>computation.
>This can improve the speed, but only if the data are much correlate, and
>this correlation you cannot explain to the query engine.
>
>If a solution not exist, you need to redesign the structure of tables.
>
>Fabio