So I know this problem isn't a new one, but I'm trying to wrap my head around it and understand the best way to deal with scenarios like this.

Say I have a hypothetical table 'X' that looks like this:

GroupID ID (identity)   SomeDateTime
--------------------------------------------
1       1000        1/1/01
1       1001        2/2/02
1       1002        3/3/03
2       1003        4/4/04
2       1004        5/5/05

I want to query it so the result set looks like this:

----------------------------------------
1       1002        3/3/03
2       1004        5/5/05

Basically what I want is the MAX SomeDateTime value grouped by my GroupID column. The kicker is that I DON'T want to group by the ID column, I just want to know the 'ID' that corresponds to the MAX SomeDateTime.

I know one pseudo-solution would be:

;WITH X1 as (
    SELECT MAX(SomeDateTime) as SomeDateTime, GroupID 
    FROM X
    GROUP BY GroupID
)
SELECT X1.SomeDateTime, X1.GroupID, X2.ID
FROM X1
    INNER JOIN X as X2
        ON X.DateTime = X2.DateTime

But this doesn't solve the fact that a DateTime might not be unique. And it seems sloppy to join on a DateTime like that.

Another pseudo-solution could be:

SELECT X.GroupID, MAX(X.ID) as ID, MAX(X.SomeDateTime) as SomeDateTime
FROM X
GROUP BY X.GroupID

But there are no guarantees that ID will actually match the row that SomeDateTime comes from.

A third less useful option might be:

SELECT TOP 1 X.GroupID, X.ID, X.SomeDateTime
FROM X
WHERE X.GroupID = 1
ORDER BY X.SomeDateTime DESC

But obviously that only works with a single, known, GroupID. I want to be able to join this result set on GroupID and/or ID.

Does anyone know of any clever solutions? Any good uses of windowing functions?

Thanks!