And wOOdy, that’s a great tip!

UPDATE Orders ao ;
SET ao.cust_key = ;
(SELECT MAX(ac.cust_key) FROM Customer ac ;
WHERE ac.cust_id = ao.cust_id ;
GROUP BY ac.cust_id, ac.cust_fname, ac.cust_lname, ac.cust_phone)

INDEX ON [expression for doublicate fields] UNIQUE TO temp.idx

Thus, based on your sample code it would be:
INDEX ON STR(cust_id) + cust_fname+ cust_lname+ cust_phone UNIQUE TO temp

SET DELETED OFF
USE Customers
DELETE ALL
INDEX ON UNIQUE TO _Temp
RECALL ALL
SET ORDER TO
SET DELETED ON

The trick here is the use of an UNIQUE (xBase style, not SQL) indextype, which always only adds the very first occurence into the index. This way the FIRST entry survives (which is usually the correct one) and subsequent entries are ignored.
Based on that index the RECALL activates only those records listed in the index. After that, this UNIQUE-index can get deleted.

If a table with duplicates has children, Brian’s code should be run first to re-assign all keys to the most current (the MAX) record.

Thanks for the input!

For example:

select distinct (cust_fname+cust_lname) as x from Customer where (cust_fname+cust_lname) in ( select (cust_fname+cust_lname) as x from Customer group by x having count(*) > 1)

Returns two records:
KevinRagsdale
RickStrahl

But doesn’t tell me which one I can delete. Unless, of course I can identify an attribute that is consistently different. If I change the above select to a DELETE, every occurrence of KevinRagsdale and RickStrahl will be deleted.

Actually, come to think of it, the method outlined in this article is best suited for cases when there is a small difference in the record. I see this sometimes with a timestamp or surrogate key. If rows are identical in every way, then it doesn’t matter. The trick is to identify which attribute is the “anchor” and use that in the WHERE clause.

Another way around this: you could retrofit a primary key by using RECNO or some other method on the Customer entity:

ALTER TABLE Customer ADD COLUMN retrofit integer
REPLACE ALL retrofit WITH RECNO() IN Customer

And then simply use new attribute “retrofit” in the WHERE and MAX clauses.

select distinct([attribute1+attribute2]) as x from [tablename] where ([attribute1+attribute2]) in ( select ([attribute1+attribute2]) as x from [tablename] group by x having count(*) > 1)

How do you handle child records linked on cust_key, perhaps a similar update query? Untested code follows:

UPDATE Customer ac ;
SET cust_key = ;
(SELECT MAX(cust_key) FROM Customer bc;
WHERE bc.cust_id = ac.cust_id ;
GROUP BY cust_id, cust_fname, cust_lname, cust_phone)