To do a legitimate test on this kind of issue you really have to reboot between attempts.

If you don't then you can (and will) have the OS cache the table data involved so any timings after the first are not reliable.

While it may be that you *have* identified some performance connection between SET EXACT and a query, the design intention of VFP is that SET ANSI is the one applicable to SQL queries and SET EXACT is for all ELSE.


>Tamar,
>I'm not too sure about your last statement.
>I just tested a table with 10k records with the following query from the command window:-
>sele invoice_no, invoice_date from inv_table order by 1 into cursor c_t
>Ran the query multiple times, each time shutting down VFP and restarting.
>One test involve set exact off preceding the query and the other involve exact on.
>It surprised me to find that on average, the one with set exact off averaged < 0.5 seconds while the one with exact on averaged almost 2 seconds to complete.
>Its done on VFP6SP5 on a Pentium III notebook with 64MB Ram.
>Its got indexes on the invoice_no field as well as a !deleted() tag.
>I don't really know how to explain this.
>Yau