Update Only One of Duplicate Records
December 5, 2007 Hey, Ted:
|
We have a file that has no unique key. This means that we sometimes get duplicate records in the file. This does not cause us problems in RPG programs because we can read a record, make changes to it, and update it, whether we’re using native I/O or an SQL cursor. However, having duplicate rows is a problem when we use SQL outside of programs. If two or more records are identical, there is no way to code a WHERE clause that will select only one record. Or is there? –David It is ugly, but possible, David. Assume table MYDATA, containing the following information: Name Age Location ============ === ======== Bill Fold 18 Moscow Sue Doe-Nymm 25 Athens Ben Dover 22 Helsinki Bill Fold 18 Moscow Polly Wogg 19 Helsinki Notice that the first and fourth rows are identical. Here’s the UPDATE command to change the name in only one of them.
UPDATE qtemp.mydata AS a
SET NAME = 'Sam O''Var'
WHERE rrn(a) =
(SELECT MAX(rrn(b))
FROM qtemp.mydata AS b
WHERE NAME = 'Bill Fold')
Let’s see what’s taking place. Notice first the subquery in the last three lines. It returns the relative record number of the last row with Bill Fold in the name column. (If you prefer to update the first of the duplicate rows, use the MIN function instead.) The UPDATE command changes the name in the row whose relative record number matches the relative record number returned in the subquery. Voilà! One, and only one, row updated. Name Age Location ============ === ======== Bill Fold 18 Moscow Sue Doe-Nymm 25 Athens Ben Dover 22 Helsinki Sam O'Var 18 Moscow Polly Wogg 19 Helsinki By the way, this discussion points out yet another difference between the positioned update and the searched update, which I examined in Updating Through an SQL Cursor.
|
