SQL values disappear when using max dates - sql

Question

First time posting here and have a query that I hope someone maybe able to help with, i have tried to search for the answer but with no joy.
When i use the below SQL to find a value (in this case eb.annualvalue) it returns multiple values because no end dates have been entered into the eb table and there are too many employees without end dates for me to close down.
LEFT JOIN
(
SELECT
eb.empid, eb.bencode, eb.currencycode AS [currencycode], eb.notes AS [notes], eb.annualvalue
FROM
employeebenefit AS [eb]
WHERE
eb.bencode IN ('US 401K Plan')
AND (eb.enddate IS NULL OR eb.enddate >= '20180101')
)
AS eb26
ON eb26.empid = e.empid
However, when i use MAX startdate (code below) it returns the correct number or rows however, the eb.annualvalue figure disappears.
LEFT JOIN
(
SELECT
eb.empid, eb.bencode, eb.currencycode AS [currencycode], eb.notes AS [notes], eb.annualvalue
FROM
employeebenefit AS [eb]
WHERE
eb.bencode IN ('US 401K Plan')
AND (eb.enddate IS NULL OR eb.enddate >= '20180101')
AND (eb.startdate = (SELECT MAX(eb.startdate) FROM employeebenefit AS [eb]))
)
AS eb26
ON eb26.empid = e.empid
Any help would be greatly appreciated. Thanks Dan.

Answer 1

This sounds like a greatest-n-per-group problem, you just want one row per employee, from a table with many rows per employee. I'm not 100% clear on how you want to select that one row, but I can give an example.
Ideally, you would use ROW_NUMBER() but that only came in to effect from SQL Server 2008 onward.
The two commons alternative are:
- Join on your data twice. Once to find the "highest date" per user, again to find the whole row.
- Use a correlated sub-query to work out an individual's best row (still really joining twice)
Simple-self-join:
LEFT JOIN
(
SELECT
empid,
MAX(startdate) AS max_startdate
FROM
employeebenefit
WHERE
bencode IN ('US 401K Plan')
AND (enddate IS NULL OR enddate >= '20180101')
GROUP BY
empid
)
latest_employeebenefit
ON latest_employeebenefit.empid = e.empid
LEFT JOIN
employeebenefit
ON employeebenefit.empid = latest_employeebenefit.empid
AND employeebenefit.startdate = latest_employeebenefit.max_startdate
AND employeebenefit.bencode IN ('US 401K Plan')
AND (employeebenefit.enddate IS NULL OR employeebenefit.enddate >= '20180101')
This has the "feature" that if two such records both match the max_startdate (a tie) then both will come through. Often that is impossible, often it's desirable, it depends on your data and your needs.
Correlated-sub-query for join:
LEFT JOIN
employeebenefit
ON employeebenefit.id =
(
SELECT TOP(1) lookup.id
FROM employeebenefit AS lookup
WHERE lookup.empid = e.empid -- the correlated bit
AND lookup.bencode IN ('US 401K Plan')
AND (lookup.enddate IS NULL OR lookup.enddate >= '20180101')
ORDER BY lookup.startdate DESC
)
This is slightly different in that it always returns just one row. If there can be a tie when only sorting by startdate it's generally best to add another column to the ORDER BY, even if it's just an id column, to ensure the results are deterministic.

Answer 2

You can use the code bellow , if I undestood your question
OUTER APPLY
(
SELECT TOP 1
eb.empid, eb.bencode, eb.currencycode AS [currencycode], eb.notes AS [notes], eb.annualvalue
FROM
employeebenefit AS [eb]
WHERE
eb.empid = e.empid
AND eb.bencode IN ('US 401K Plan')
AND (eb.enddate IS NULL OR eb.enddate >= '20180101')
ORDER BY
eb.startdate DESC
)
AS eb26