Sergey,
>>Hilmar,
>>SNIP
>>>
>>>An index on strtran(...) is virtually useless, since it is of variable length. Any decent VFP index has a fixed length. Perhaps padr(strtran(...)) will achieve better results.
>>
>>I've heard this many many times, and every time that I read it I wonder to myself...
>>1) Don't all VFP indexes have a fixed length anyway, INTERNALLY?
>>2) If VFP lets one create such a TAG, why would it then be useless???
>>
>>Perhaps you mean that it won't be used by Rushmore?
>>
>>Jim
>
>This index will be used by Rushmore. However, it isn't relable because the lenght of the index key depends on the record VFP uses to determine it. If table is empty, you wouldn't be able to create such index at all.

Your second point makes the case for me!

As regards the length of the index key... does INDEX ON operate the same way that SQL-Select does as to the way SQL uses the first (qualifying?) field's content to establish the length of all output of the field? It seems most unlikely to me that INDEX ON would operate that way.

thanks