Sorry, you're right, I misunderstood your procedure. It should work on my project. But since I don't have time to do any more benchmarking and that David idea is more than fast enough for my need, I will stick with it for now.

I will keep your idea in mind for the next time. Thanks for your help.

>Unless I've completely misunderstood, the same logic applies to your SDF file.
>testtim SDF file = 2505 records with 1490 non-duplicate records, in other words, the file has 1015 duplicate records in the SDF file.
>
>Code :
>USE c:\brad\testtim3 ORDER telephone EXCL
>USE c:\brad\testtim2 IN 0 EXCL
>lctime = SECONDS()
>SELECT testtim2
>APPEND FROM c:\brad\testtim.txt TYPE SDF
>SET RELA TO telephone INTO testtim3
>SCAN
> IF !FOUND('testtim3')
> INSERT INTO testtim3 ;
> (telephone) ;
> VALUES ;
> (testtim2.telephone)
> ENDIF
>ENDSCAN
>
>?SECONDS() - lctime
>
>After running the above code on a P166 with VFP5.0a, time taken was .4 to .5 seconds to process 2500 records and to have a table with no duplicate numbers in it.