I've been thinking about that, and I think you can do it in one of 2 ways, though one is not that elegant.

First method (preferable):

=RAND(-1)
COPY FIELDS primarykey TO temp FOR RAND()<0.10
SELECT * FROM mytable ;
WHERE primarykey IN ;
(SELECT primarykey FROM temp) ;
INTO CURSOR result
ERASE temp.dbf


Second method (clumsy but interesting):

=RAND(-1)
SELECT * FROM mytable ;
WHERE TenPctSample() ;
INTO CURSOR result

FUNCTION TenPctSample
RETURN RAND()<0.10


The problem with doing a SELECT...WHERE RAND()<0.10 is that VFP evaluates the WHERE clause before it even executes it. Therefore, it would evaluate the WHERE clause up-front to be .T. only 10% of the time and .F. 90% of the time; therefore, the SELECT would return NOTHING 90% of the time. However, VFP won't attempt to evaluate a user-defined function up-front, thus the second method above works. It's not that straightforward, and evaulating a user-defined function can slow things down.

Actually, I just thought of a 3rd way, but I don't like it performance-wise on a huge table:

SELECT *,RAND()<0.10 AS itsok ;
FROM mytable ;
HAVING itsok=.T. ;
INTO CURSOR result


The above will essentially build a copy of the ENTIRE table with an extra field tacked on, and then the HAVING clause will strip out the 90% of the records you don't want. It works, but it entails a lot of overhead if you're working on a huge table.

--Brad