>>>
>>>I think Mike hit on the LOCATE delay -- it's not long enough to build an index on a large table (and where the index definitely exists), just a one or two second pause...
>>
>>It could be that for a LOCATE, that if an index is not found that it just does a sequential search. I know that in a SQL SELECT, Rushmore will sometimes build an index and sometimes not...it tries to figure out if it needs one.
>
>But in this case, the initial LOCATE has a pause (it seems to be doing a rather slow search for the index, like it's considering the options, then zips to the record), and then subsequent CONTinues are instantaneous...and we're talking major table size here...well, I always use SEEKs programmatically anyway on these tables, so it's kind of moot that LOCATE may be a bit slower...

OK...sounds like it is looking for an index it can use, then once it finds it, it is used on the CONT.