Question

I have the following fields: user_id, app_mod, profile, update_date.
The first three fields can be duplicated (i.e. two rows could have user_id, app_mod and profile equals, but different update_date):

user_id   app_mod  profile update_date
560047    RI03290   22809   14-GEN-10  
560047    RI03290   22809   18-GEN-10

How can I get the one row with higher update_date in an Oracle environment?

560047    RI03290     22809   18-GEN-10

Answer 1

If the table really only has 4 columns

SELECT user_id, app_mod, profile, max(update_date)
  FROM table_name
 GROUP BY user_id, app_mod, profile

If the table has other columns as well that you’re not including and that will not be duplicated that you do want returned

SELECT *
  FROM (SELECT a.*,
               rank() over (partition by user_id, app_mod, profile
                                order by update_date desc) rnk
          FROM table_name a)
 WHERE rnk = 1;

The analytic function rank will assign multiple rows the same rank if there are ties (i.e. the two rows have the same user_id, app_mod, profile, and update_date). You could use row_number instead which would arbitrarily break the tie. But that would not generally be a repeatable process– Oracle might pick one of the two rows today and another row tomorrow. If you were looking for rows other than the most recent row, you might also look at dense_rank but that will behave identically to rank if you’re just looking for the most recent row.