SQL: Delete all NOT MAX Records in GroupBy - sql

Question

My goal is to delete all records from my table that are NOT the MAX(recordDate) of a grouped CaseKey. So if I have 9 records with 3 sets of 3 casekeys, and each casekey has its 3 dates. I'd delete the 2 lower dates of each set and come up with 3 total records, only the MAX(recordDate) of each remaining.
I have the following SQL Query:
DELETE FROM table
WHERE tableID NOT IN (
SELECT tableID
FROM (
Select MAX(recordDate) As myDate, tableID From table
Group By CaseKey
) As foo
)
I receive the error:
Error on Line 3... Column 'table.tableID' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.
Obviously I could add tableID to my Group By clause, but then the result of that statement is incorrect and returns all rows instead of just returning the MAX recordDate of the grouped CaseKeys.
Server is down right now, but the apparent answer is: (tiny tweak from WildPlasser's answer)
DELETE zt FROM ztable zt
WHERE EXISTS (
SELECT * FROM ztable ex
WHERE ex.CaseKey = zt.CaseKey
AND ex.recordDate > zt.recordDate
);
In other words, for each record in zt, run a query to see if the same record also has a record with a higher recordDate. If so, the WHERE EXISTS statement passes and the record is deleted, otherwise the WHERE statement fails and the record is its own MAX recordDate.
Thank you, WildPlasser, for that simplistic methodology that I was somehow blowing up.

Answer 1

There is one special property of MAX: there is no record with a higher value than max. So we can delete all the records for which a record with the same CaseKey, but with a higher recordDate exists:
DELETE FROM ztable zt
WHERE EXISTS (
SELECT *
FROM ztable ex
WHERE ex.CaseKey = zt.CaseKey
AND ex.recordDate > zt.recordDate
);
BTW: The above query (as well as the MAX() version) assumes that there is only one record with the maximum date. There could be ties.
In the case of ties, you'll need to add an extra field to the where clause; as a tie-breaker. Assuming that TableId can function as such, the query would become:
DELETE FROM ztable zt
WHERE EXISTS (
SELECT *
FROM ztable ex
WHERE ex.CaseKey = zt.CaseKey
AND ( ex.recordDate > zt.recordDate
OR (ex.recordDate = zt.recordDate AND ex.TableId > zt.TableId)
)
);

Answer 2

Just express
delete all records from my table that are NOT the MAX(recordDate) of a
grouped CaseKey
in sql as
DELETE FROM table t1
WHERE t1.recordDate <>
(SELECT MAX(recordDate)
FROM table t2
WHERE t2.CaseKey = t1.CaseKey)

Answer 3

You can rank all records with the same caseKey where the rank > 1 to only return the lower dates. That way you can use your tableID.
DELETE FROM [table]
WHERE [tableID] IN
(SELECT
[sub].[tableID]
FROM
(
SELECT
[tableID],
Rank() OVER (PARTITION BY [caseKey] ORDER BY [recordDate] DESC, [tableID] DESC) AS [rank]
FROM [table]
) AS [sub]
WHERE [sub].[rank] > 1)