Hi brad,

This behaviour is documented in the VFP9 helpfile.
It is an enhancement to define uniqueness and still beeing rushmore optimizable, where in pre VFP9 we had to have two indexes. In technical terms this enhancement is not as stunning. In order to perform optimization it will fully load the index tag to determine the non deleted records and the nodes that correspond to the index expression. This means that such indexes perform way less as opposed to having a seperate addtinal index. I/O is just way less in the latter and this will esspecially visible in large tables.

All in all, I wish they had not implemented this feature at all. The hidden performance downside might hurt a lot of implementations and let their creators wondering what the hell is going on..

Walter,