Finding a match for an entry on a date - sql

Question

I am looking for a query that would find all entries that have a login without a logout.
My data looks like this
Key Date Employee
LOGIN 20171225 111
LOGIN 20171225 111
LOGIN 20171226 111
There should be a record here. I need to catch that.
LOGIN 20171227 111
LOGIN 20171227 111
12345 20171227 222 (There is also a LOT of other random data in the table.)
Select Date, Employee
From My Table
Where Key = 'LOGIN'
Group by date, employee
Order by employee
I don't know how to filter out to see if there is one or two logins for that day. I'll need to see where there's only one, because that indicates they have not logged out. This isn't giving me the correct information.
Thank you.

Answer 1

You might use this
DECLARE #dummyTbl TABLE([Key] VARCHAR(100),[Date] DATE,Employee INT);
INSERT INTO #dummyTbl VALUES
('LOGIN','20171225',111)
,('LOGIN','20171225',111)
,('LOGIN','20171226',111)
,('LOGIN','20171227',111)
,('LOGIN','20171227',111);
SELECT *
FROM #dummyTbl
GROUP BY [Key],[Date],Employee
HAVING COUNT(*)=1
But I wonder, why your Key is LOGIN in both cases, why not use LOGOUT?

Answer 2

If your key is always LogIn then you really want to look for any ODD number of entries. Which you can do by using remainder (%) of division by 2 not equal to 0.
DECLARE #dummyTbl TABLE([Key] VARCHAR(100),[Date] DATE,Employee INT);
INSERT INTO #dummyTbl VALUES
('LOGIN','20171225',111)
,('LOGIN','20171225',111)
,('LOGIN','20171226',111)
,('LOGIN','20171227',111)
,('LOGIN','20171227',111);
SELECT *
FROM #dummyTbl
GROUP BY [Key],[Date],Employee
HAVING COUNT(*) % 2 <> 0
If you have multiple keys login/logout and you are attempting to figure out if the user is Logged Out then it is best to look at the last value for the user for the day and if it is not logout then you know they are still logged in.
DECLARE #dummyTbl TABLE(Id INT IDENTITY(1,1), [Key] VARCHAR(100),[Date] DATE,Employee INT);
INSERT INTO #dummyTbl VALUES
('LOGIN','20171225',111)
,('LOGIN','20171225',111)
,('LOGOUT','20171225',111)
,('LOGIN','20171226',111)
,('LOGIN','20171227',111)
,('LOGOUT','20171227',111);
;WITH cteRowNum AS (
SELECT *
,LastDailyActivityRowNum = ROW_NUMBER() OVER (PARTITION BY Date, Employee ORDER BY Id DESC)
FROM
#dummyTbl
)
SELECT *
FROM
cteRowNum
WHERE
LastDailyActivityRowNum = 1
AND [Key] = 'LOGIN'
If you potentially have dirty data (missing login or logout record) then it gets a bit more complicated where you will have to make some business decisions but the last record method is still generally the way to go. When you have employees that can work past midnight without logging out then it gets a bit more complicated too.....