Question

I have a table, with rowid, userid, productid, and times. once I ordered the table by userid and time. It looks like:

u1, t1, p1
u1, t2, p1
u1, t3, p1
u1, t4, p2
u1, t5, p2
u1, t6, p3
u2, t7, p1
u2, t8, p1
u2, t9, p2
u2, t10,p3
 .....

I want to delete those rows whose product id is same as the previous row.
therefore the final table should be :

u1, t1, p1
u1, t4, p2
u1, t6, p3
u2, t7, p1
u2, t9, p2
u2, t10,p3
.....

How can I do this in SQL? Many thanks

Answer 1

^{Simplified version}
Turns out it is for SQL Server, and you can delete from the CTE directly in SQL Server (as hinted by @Royi Namir in the comments below):

;WITH x AS (
    SELECT rowid,
           rownumber() OVER (PARTITION BY user_id,productid ORDER BY times) AS rn
    FROM   tbl
    )
DELETE FROM x
WHERE  rn > 1;

Simplified live demo at sqlfiddle.
Much like this one.