>>>Still - I don't see the need of the optimizing behavior vfp shows.
>>
>>I'd agree with that, except that if VFP optimized UDFs, there would be a ton of complaints where someone's UDF was misbehaving and they'd blame VFP and call it buggy.
>
>I am not following you: If I CREATE a UDF based index und use it via seek/scan the result it gets used instantly, if the index expression is "seen" in a join, a temporary index is created on the same expression and the UDF momentarily loaded. If the UDF is the same, the result will be the same as if the information in the cdx is reused. There also is the re-evaluation of each clause after a rushmored query to make sure no false records are loaded. And anybody stupid enough to load a different UDF regularly will certainly be bold enough to sound off about a buggy VFP <bg>.

The UDF has to return fixed length values. If a programmer does not know that, the results will be strange and they will blame vfp.

VFP can at least try to compute length based on known functions, but how to guess what comes from UDF?