>That's logical alright, Hilmar, but by then the records themselves
should all be in cache and the time to check the deleted flag should be measured in milliseconds in total (assuming a PC in the 500mhz range or better).
>The time difference Dan reported still seems to be a real stretch to me.
>
>JimN
Well, I don't know. There is probably part of the optimization process I don't understand.
One question I have, for instance, about the optimization process, has to do with creating bitmaps (i.e., one bit for each record) in memory, and applying BITAND to them. I think we may safely assume that Visual FoxPro actually does something similar. But: does Visual FoxPro create this bitmap for a single condition? Or for a partial optimization, where only a single condition can be optimized from an index? It wouldn't seem necessary to do so, but I don't know the actual internal workings of VFP.
This question may, or may not, apply to the case on hand.
Hilmar.
Difference in opinions hath cost many millions of lives: for instance, whether flesh be bread, or bread be flesh; whether whistling be a vice or a virtue; whether it be better to kiss a post, or throw it into the fire... (from Gulliver's Travels)