SQL select case group by - sql

Question

I have a table 'LIST_USERS'.
Table Description -
USER_ID NUMBER(8)
LOGIN_ID VARCHAR2(8)
CREATE_DATE TIMESTAMP(6)
LOGIN_DATE TIMESTAMP(6)
Table data -
USER_ID LOGIN_ID CREATE_DATE LOGIN_DATE
---------------------------------------------------
101 test1 04/24/2016 null
102 test1 04/24/2016 04/29/2016
103 test2 04/25/2016 null
104 test2 04/26/2016 null
105 test3 04/27/2016 04/28/2016
106 test3 04/27/2016 04/29/2016
107 test4 04/28/2016 04/29/2016
987 test5 04/29/2016 null
109 test5 04/29/2016 null
108 test5 04/29/2016 04/29/2016
Condition - I need to fetch USER_ID, and LOGIN_ID from 'LIST_USERS' table based of max LOGIN_DATE. If LOGIN_DATE is null, I need to get the record based on max CREATE_DATE.
I need to get the below result -
USER_ID LOGIN_ID
---------------------
102 test1
104 test2
106 test3
107 test4
108 test5
I am using the below query. But it will give me only LOGIN_ID, and 'Login_Or_Create_Date' but I need USER_ID, and LOGIN_ID. Is there way I can get USER_ID as well as in the result shown above?
select LOGIN_ID,
(case when max(LOGIN_DATE) is null then max(CREATE_DATE)
else max(LOGIN_DATE) end) as Login_Or_Create_Date
from LIST_USERS;

Answer 1

Try this:
SELECT USER_ID, LOGIN_ID
FROM (
SELECT USER_ID, LOGIN_ID,
ROW_NUMBER() OVER (PARTITION BY LOGIN_ID
ORDER BY COALESCE(LOGIN_DATE, CREATE_DATE) DESC) AS rn
FROM LIST_USERS) t
WHERE t.rn = 1

Answer 2

Sounds like a job for keep dense_rank:
select min(user_id) keep (dense_rank last order by coalesce(login_date, create_date))
as user_id,
login_id
from list_users
group by login_id
order by user_id;
The last keeps the record with the latest login/create date; the coalesce() takes the login date first and falls back to the create date if that is null (or you could use nvl() instead of course). You could also do first and order by desc - the result is the same (if there are no nulls anyway, and it looks like there shouldn't be), but last feels more intuitive when you want the latest date I think.
Demo using your data in a CTE:
with list_users(user_id, login_id, create_date, login_date) as (
select 101, 'test1', date '2016-04-24', null from dual
union all select 102, 'test1', date '2016-04-24', date '2016-04-29' from dual
union all select 103, 'test2', date '2016-04-25', null from dual
union all select 104, 'test2', date '2016-04-26', null from dual
union all select 105, 'test3', date '2016-04-27', date '2016-04-28' from dual
union all select 106, 'test3', date '2016-04-27', date '2016-04-29' from dual
union all select 107, 'test4', date '2016-04-28', date '2016-04-29' from dual
)
select min(user_id) keep (dense_rank last order by coalesce(login_date, create_date))
as user_id,
login_id
from list_users
group by login_id
order by user_id;
USER_ID LOGIN
---------- -----
102 test1
104 test2
106 test3
107 test4
And with your modified data:
with list_users(user_id, login_id, create_date, login_date) as (
select 101, 'test1', date '2016-04-24', null from dual
union all select 102, 'test1', date '2016-04-24', date '2016-04-29' from dual
union all select 103, 'test2', date '2016-04-25', null from dual
union all select 104, 'test2', date '2016-04-26', null from dual
union all select 105, 'test3', date '2016-04-27', date '2016-04-28' from dual
union all select 106, 'test3', date '2016-04-27', date '2016-04-29' from dual
union all select 107, 'test4', date '2016-04-28', date '2016-04-29' from dual
union all select 987, 'test5', date '2016-04-29', null from dual
union all select 109, 'test5', date '2016-04-29', null from dual
union all select 108, 'test5', date '2016-04-29', date '2016-04-29' from dual
)
select min(user_id) keep (dense_rank last order by coalesce(login_date, create_date))
as user_id,
login_id
from list_users
group by login_id
order by user_id;
USER_ID LOGIN
---------- -----
102 test1
104 test2
106 test3
107 test4
108 test5