>Snip..
>>>You're right but "go top" with a filter is so slow. Use locate instead.
>>>cetin
>>You could be right about that. I ran some tests(100 reps of each) on a 400K record table. I haven't had time to really go thru the figures but it looks like on a random filter set on a EXCLU open the goto ran a little faster. With the same test on a SHARE open the locate ran faster(but both took about twice the time as the EXCLU open). I'll run the 'NO RECORDS' filter later - they take about 3 minutes to run 10 tests(about 12 times the random filter shared) and I want to do 100 tests on each type open. I'll let you know what I find out.
>Sammie,
>To cut down the tests a little, add a dummy index tag for deleted or nondeleted(index on !deleted() tag dummy) and delete a lot. Run the test with fully, partially optimizable expressions. I already tested them for years and it's the pros not only me, indicating to the importance of speed with locate vs gotop-gobottom. Also try this :

set order to tag anytag
>goto 200000
>go top        && Measure
>go bottom     && Measure
>* Versus
>set order to tag anytag
>goto 200000
>* Go top
>locate        && Measure
>* Go bottom
>set order to tag anytag descending && Start measure - index upside down
>locate
>set order to tag anytag            && Stop measure 

Cetin
Cetin,
I thought YOU were one of the "PROS". I'll try what you said. Sometimes I just like to see things for myself. That's what I really like about UT - everyday I learn something. I've been hacking with Foxpro for about 6 years but the rapid changes from 2.0 and 2.5 to VFP just amaze me. The places I have worked have always concentrated on production. All kinds of money for equipment and software - nothing for training. Would you believe that the last "formal" training I had was for dBase 2.0 on a dual floppy machine. But the company I work for has finally agreed to send me to Devcon. Maybe there is hope after all.