>
>Well, for me it's not so interesting whether a search takes 0.01 millisecond or 0.02 milliseconds.....! I have timed real life data searches before and after a reindex, and the difference is minimal. Given enough RAM and fast hardware, my conclusion is that a reindex is practically a waste of time. But "your milage may vary".

Agreed - especially the last sentence. My case is insofar untypical as the disk cache does not always encompass all the cdx in a 32-Bit OS and the times for one simple seek operation on between(reccount(), 5*10**6 and 3*10**7) tend to cluster in the 1 - 6 ms range. Running a couple of such seeks each more than 1.4*10**5 times will be felt - when all reindex operations together need about 5 minutes I get a marked difference. One of the "drawbacks" of vfp is that whenever something cannot be handled "automatically intelligently inside" vfp perfomance on that particular line tends to drop a lot in one big step. But that makes it fun - for me at least <bg>.

regards

thomas