>>Ed,
>>
>>Just to let you know what I am trying to do:
>>
>>I have a Customer table and a child Purchased table. I want an SQL to determine those Customers who bought
>>SOAP and TOWELS and TOOTHBRUSH.
>>
>>The UDF (obviously not the one show) determines if the pair Cust_id,Item_id exists for each Item in the search list
>>(by using SEEK) and returns .T. or .F. The values of Item_id are stored in an array.
>>
>>Is this a totally off the wall approach and dangerous to boot?
>
>

>select cust_id from customer where cust_id in
>      (select cust_id from purchased
>         where item_id = soap and cust_id in
>             (select cust_id from purchased
>                where item_id = towels  and cust_id in
>                    (select cust_id from purchased where item_id = toothbrush)
>             )
>      )

I'd take a radically different approach:  build a cursor with the items that
must be on the order to qualify, and call that MustHave.  Now try:

nNumItems = RECCOUNT('MustHave')
SELECT cust_id, item_id FROM Purchased ;
  INTO CURSOR CandCusts ;
 WHERE item_id IN (SELECT item_id from MustHave) ;
 GROUP BY cust_id, Item_Id ;
 ORDER BY cust_id, Item_id

SELECT Cust_ID, COUNT(*) AS NumMatches FROM CandCusts ;
  INTO CURSOR TheseCustsBoughtEverything ;
 GROUP BY Cust_ID ;
HAVING NumMatches = nNumItems

The first SELECT pulls order items that are on our list of MustHave items; we want only one record per cust_id, item_id pair, so I've cheated and used a GROUP BY rather than a DISTINCT clause to cull duplicates. We now have a table of all cust_ids that matched some item_id we were looking for.

We want only those cust_ids that have all the MustHave items; we can discover this by counting how many records are in our intermediate candidate file for each cust_id. Any cust_id that has as many records as there are MustHave items represent customers who purchased evrything of interest at least one time.