Update using an aggregate function

Question

Hi all, while I'm a first time poster I've searched this forum as much as I can for a solution to this problem.
I'm trying to update an end_time column in a table using the values in the start_time column.
Here's a simplified example of what I want:
update table
set end_time = lag(start_time) over (order by start_time desc)-1;
So,
KEY START_TIME
1 7/1/1998
2 5/1/1998
3 4/1/1998
4 6/1/1995
becomes
KEY START_TIME END_TIME
1 7/1/1998
2 5/1/1998 6/30/1998
3 4/1/1998 4/30/1998
4 6/1/1995 3/31/1998
While I'd love this statement to be valid, does anyone know an alternative? I'd like to use the merge statement but I'm developing under an oracle 8i environment. I'm saving a pl loop as a last option.

Answer 1

It is possible that with large data sets this first solution could be slow. The following query may choose full table scans and sort merges which might be more preferred for big tables. Also, the actual updates in the following MERGE will not execute unless the values have actually changed:
MERGE INTO foo f USING
SELECT a.start_date, min(b.start_date)-1 end_date
FROM   foo a,
        foo b
WHERE  b.start_date(+) > a.start_date
GROUP BY a.start_date
) bar
ON (f.start_date = bar.start_date)
WHEN MATCHED THEN UPDATE SET f.end_date = bar.end_date
WHERE nvl(f.end_date,to_date('01-jan-1900','dd-mon-yyyy')) != nvl(bar.end_date,to_date('01-jan-1900','dd-mon-yyyy'));Greg Pike
http://www.singlequery.com