SQL First Time In Last Time Out (Transact-SQL..preferrably) - sql

Question

I have the following table which contains all the time in and time out of people:
CREATE TABLE test (
timecardid INT
, trandate DATE
, employeeid INT
, trantime TIME
, Trantype VARCHAR(1)
, Projcode VARCHAR(3)
)
The task is to get all the earliest trantime with trantype A (perhaps using MIN) and the latest trantime with trantype Z (Using Max), all of which in that trandate (ie. trantype A for july 17 is 8:00 AM and trantype Z for july 17 is 7:00PM).
the problem is, the output should be in the same format as the table where it's coming from, meaning that I have to leave this data and filter out the rest (that aren't the earliest and latest in/out for that date, per employee)
My current solution is to use two different select commands to get all earliest, then get all the latest. then combine them both.
I was wondering though, is there a much simpler, single string solution?
Thank you very much.
EDIT (I apologize, here is the sample. Server is SQL Server 2008):
Timecardid | Trandate | employeeid | trantime | trantype | Projcode
1 2013-04-01 1 8:00:00 A SAMPLE1
2 2013-04-01 1 9:00:00 A SAMPLE1
3 2013-04-01 2 7:00:00 A SAMPLE1
4 2013-04-01 2 6:59:59 A SAMPLE1
5 2013-04-01 1 17:00:00 Z SAMPLE1
6 2013-04-01 1 17:19:00 Z SAMPLE1
7 2013-04-01 2 17:00:00 Z SAMPLE1
8 2013-04-02 1 8:00:00 A SAMPLE1
9 2013-04-02 1 9:00:00 A SAMPLE1
10 2013-04-02 2 7:00:58 A SAMPLE1
11 2013-04-02 2 18:00:00 Z SAMPLE1
12 2013-04-02 2 18:00:01 Z SAMPLE1
13 2013-04-02 1 20:00:00 Z SAMPLE1
Expected Results (the earliest in and the latest out per day, per employee, in a select command):
Timecardid | Trandate | employeeid | trantime | trantype | Projcode
1 2013-04-01 1 8:00:00 A SAMPLE1
4 2013-04-01 2 6:59:59 A SAMPLE1
6 2013-04-01 1 17:19:00 Z SAMPLE1
7 2013-04-01 2 17:00:00 Z SAMPLE1
8 2013-04-02 1 8:00:00 A SAMPLE1
10 2013-04-02 2 7:00:58 A SAMPLE1
12 2013-04-02 2 18:00:01 Z SAMPLE1
13 2013-04-02 1 20:00:00 Z SAMPLE1
Thank you very much

Answer 1

Perhaps this is what you're looking for:
select
t.*
from
test t
where
trantime in (
(select min(trantime) from test t1 where t1.trandate = t.trandate and trantype = 'A'),
(select max(trantime) from test t2 where t2.trandate = t.trandate and trantype = 'Z')
)
Changing my answer to account for the "per employee" requirement:
;WITH EarliestIn AS
(
SELECT trandate, employeeid, min(trantime) AS EarliestTimeIn
FROM test
WHERE trantype = 'A'
GROUP BY trandate, employeeid
),
LatestOut AS
(
SELECT trandate, employeeid, max(trantime) AS LatestTimeOut
FROM test
WHERE trantype = 'Z'
GROUP BY trandate, employeeid
)
SELECT *
FROM test t
WHERE
EXISTS (SELECT * FROM EarliestIn WHERE t.trandate = EarliestIn.trandate AND t.employeeid = EarliestIn.employeeid AND t.trantime = EarliestIn.EarliestTimeIn)
OR EXISTS (SELECT * FROM LatestOut WHERE t.trandate = LatestOut.trandate AND t.employeeid = LatestOut.employeeid AND t.trantime = LatestOut.LatestTimeOut)

Answer 2

Assuming timecardid column is PK or unique, and if I understand it correctly, I would do something like
DECLARE #date DATE
SET #date = '2013-07-01'
SELECT
T0.*
FROM
(SELECT DISTINCT employeeid FROM test) E
CROSS APPLY (
SELECT TOP 1
T.timecardid
FROM
test T
WHERE
T.trandate = #date
AND T.Trantype = 'A'
AND T.employeeid = E.employeeid
ORDER BY T.trantime
UNION ALL
SELECT TOP 1
T.timecardid
FROM
test T
WHERE
T.trandate = #date
AND T.Trantype = 'Z'
AND T.employeeid = E.employeeid
ORDER BY T.trantime DESC
) V
JOIN test T0 ON T0.timecardid = V.timecardid
Appropriate indexes should be set for the table, if you aware of performance.

Answer 3

If you're using SQL server 2012, you can use LAG/LEAD to find the max and min rows in a fairly concise way;
WITH cte AS (
SELECT *,
LAG(timecardid) OVER (PARTITION BY trandate,employeeid,trantype ORDER BY trantime) lagid,
LEAD(timecardid) OVER (PARTITION BY trandate,employeeid,trantype ORDER BY trantime) leadid
FROM test
)
SELECT timecardid,trandate,employeeid,trantime,trantype,projcode
FROM cte
WHERE trantype='A' AND lagid IS NULL
OR trantype='Z' AND leadid IS NULL;
An SQLfiddle to test with.

Answer 4

I would use ROW_NUMBER to sort out the rows you want to select:
;with Ordered as (
select *,
ROW_NUMBER() OVER (PARTITION BY Trandate,employeeid,trantype
ORDER BY trantime ASC) as rnEarly,
ROW_NUMBER() OVER (PARTITION BY Trandate,employeeid,trantype
ORDER BY trantime DESC) as rnLate
from
Test
)
select * from Ordered
where
(rnEarly = 1 and trantype='A') or
(rnLate = 1 and trantype='Z')
order by TimecardId
(SQLFiddle)
It produces the results you've requested, and I think it's quite readable. The reason that trantype is included in the PARTITION BY clauses is so that A and Z values receive separate numbering.