>In both cases, VFP gets the index keys for almost all records (??? this depends on the data in your tables).
>
>It would seem that in case (A), VFP has more work to do. Once more, full optimization is slower. The situation might change radically if you killed the index on DELETED().
>
>I suggest you re-run the tests without this offensive (for me) index (if you haven't done so already). I also suggest that you do the test under real-world conditions, i.e., presummably, over the network.
>
>Hilmar.


One other thing I noticed was that if the total number of unique items from a group by was relatively small, if you created an index on that field, it speeds up the group by considerably. I've never seen anything about that. Have you experienced the same thing?