SQL Server - Select Distinct of two columns, where the distinct column selected has a maximum value based on two other columns - sql

Question

I have 2 tables - TC and T, with columns specified below. TC maps to T on column T_ID.
TC
----
T_ID,
TC_ID
T
-----
T_ID,
V_ID,
Datetime,
Count
My current result set is:
V_ID TC_ID Datetime Count
----|-----|------------|--------|
2 | 1 | 2013-09-26 | 450600 |
2 | 1 | 2013-12-09 | 14700 |
2 | 1 | 2014-01-22 | 15000 |
2 | 1 | 2014-01-22 | 15000 |
2 | 1 | 2014-01-22 | 7500 |
4 | 1 | 2014-01-22 | 1000 |
4 | 1 | 2013-12-05 | 0 |
4 | 2 | 2013-12-05 | 0 |
Using the following query:
select T.V_ID,
TC.TC_ID,
T.Datetime,
T.Count
from T
inner join TC
on TC.T_ID = T.T_ID
Result set I want:
V_ID TC_ID Datetime Count
----|-----|------------|--------|
2 | 1 | 2014-01-22 | 15000 |
4 | 1 | 2014-01-22 | 1000 |
4 | 2 | 2013-12-05 | 0 |
I want to write a query to select each distinct V_ID + TC_ID combination, but only with the maximum datetime, and for that datetime the maximum count. E.g. for the distinct combination of V_ID = 2 and TC_ID = 1, '2014-01-22' is the maximum datetime, and for that datetime, 15000 is the maximum count, so select this record for the new table. Any ideas? I don't know if this is too ambitious for a query and I should just handle the result set in code instead.

Answer 1

One method uses row_number():
select v_id, tc_id, datetime, count
from (select T.V_ID, TC.TC_ID, T.Datetime, T.Count,
row_number() over (partition by t.V_ID, tc.tc_id
order by datetime desc, count desc
) as seqnum
from t join
tc
on tc.t_id = t._id
) tt
where seqnum = 1;
The only issue is that some rows have the same maximum datetime value. SQL tables represent unordered sets, so there is no way to determine which is really the maximum -- unless the datetime really has a time component or another column specifies the ordering within a day.

Answer 2

It is possible to solve this using CTEs. First, extracting the data from your query. Second, get the maxdates. Third, get the highest count for each maxdate.:
;WITH Dataset AS
(
select T.V_ID,
TC.TC_ID,
T.[Datetime],
T.[Count]
from T
inner join TC
on TC.T_ID = T._ID
),
MaxDates AS
(
SELECT V_ID, TC_ID, MAX(t.[Datetime]) AS MaxDate
FROM Dataset t
GROUP BY t.V_ID, t.TC_ID
)
SELECT t.V_ID, t.TC_ID, t.[Datetime], MAX(t.[Count]) AS [Count]
FROM Dataset t
INNER JOIN MaxDates m ON t.V_ID = m.V_ID AND t.TC_ID = m.TC_ID AND m.MaxDate = t.[Datetime]
GROUP BY t.V_ID, t.TC_ID, t.[Datetime]

Answer 3

Just to keep it simple:
You need to group by T.V_ID,TC.TC_ID,
with selecting the max of date and then to get the maximum count, you must use a sub query as follows,
select T.V_ID,
TC.TC_ID,
max(T.Datetime) as Date_Time,
(select max(Count) from T as tb where v_ID = T.v_ID and DateTime = max(T.DateTime)) as Count
from T
inner join TC
on TC.T_ID = T._ID
group by T.V_ID,TC.TC_ID,