Delete rows with continuous dates within date period in SQL Server [duplicate] - sql

Question

This question already exists:
Closed 10 years ago.
Possible Duplicate:
Trying to consolidate employer records who are continuously work for same department
I am trying to consolidate employees records who have been continuously (anything < 45 days) enrolled with the specific department
Note: If the date diff (between emp_eff_to_date and next row emp_eff_from_date) is less than 45 days then it is considered as continuous
INPUT:
EMP_ID + DEPT_ID + EMP_EFF_FROM_DATE + EMP_EFF_TO_DATE
-----------------------------------------------------------------------
10 10001 8/1/2008 10/31/2009
10 10001 11/1/2009 2/25/2010
10 10001 2/26/2010 5/1/2011
10 10001 8/1/2011 10/30/2011
10 10001 12/1/2011 10/31/2012
10 10003 7/1/2007 10/31/2007
10 10004 9/27/2004 6/8/2006
10 10004 6/30/2006 6/29/2007
10 10007 6/25/2006 6/20/2007
10 10007 8/25/2007 5/25/2008
Output desired:
EMP_ID DEPT_ID EMP_EFF_FROM_DATE EMP_EFF_TO_DATE
-------------------------------------------------------------------------
10 10001 2008-08-01 2011-05-01
10 10001 2011-08-01 2012-10-31
10 10003 2007-07-01 2007-10-31
10 10004 2004-09-27 2007-06-29
10 10007 2006-06-25 2007-06-20
10 10007 2007-08-25 2007-06-29

Answer 1

I had to do a very similar thing recently, and my first thought was a Recursive table expression, which works, but may not be the best solution depending on the amount of data that is in your table.
It is not clear whether you want to actually delete the rows from the database, or just view the results as required based on the records as they currently are.
SOLUTION 1 (SQL Fiddle)
This uses the CTE to just select the results. It will essentially find the next row where the from date is within 45 days of the current row's to date, and keep looping until there are no matches. Once done it finds the result for the latest result for each from date (MaxRecursion field), and excludes then all other rows that fall within the date range of that row.
WITH CTE AS
( SELECT *, [Recursion] = 0
FROM T
UNION ALL
SELECT T.EMP_ID,
T.DEPT_ID,
T.EMP_EFF_FROM_DATE,
T2.EMP_EFF_TO_DATE,
T.[Recursion] + 1
FROM CTE T
INNER JOIN T T2
ON T.EMP_ID = T.EMP_ID
AND T.DEPT_ID = T2.DEPT_ID
AND T2.EMP_EFF_FROM_DATE > T.EMP_EFF_FROM_DATE
AND T2.EMP_EFF_TO_DATE > T.EMP_EFF_TO_DATE
AND T2.EMP_EFF_FROM_DATE <= DATEADD(DAY, 45, T.EMP_EFF_TO_DATE)
), CTE2 AS
( SELECT *,
[MaxRecursion] = MAX(Recursion) OVER(PARTITION BY EMP_ID, DEPT_ID, EMP_EFF_FROM_DATE)
FROM CTE
)
SELECT T.EMP_ID,
T.DEPT_ID,
T.EMP_EFF_FROM_DATE,
T.EMP_EFF_TO_DATE
FROM CTE2 T
WHERE Recursion = MaxRecursion
AND NOT EXISTS
( SELECT 1
FROM CTE2 T2
WHERE T.EMP_ID = T2.EMP_ID
AND T.DEPT_ID = T2.DEPT_ID
AND T.EMP_EFF_FROM_DATE < T2.EMP_EFF_FROM_DATE
AND T.EMP_EFF_TO_DATE >= T2.EMP_EFF_TO_DATE
)
ORDER BY EMP_ID, DEPT_ID, EMP_EFF_FROM_DATE, EMP_EFF_TO_DATE;
SOLUTION 2 (SQL Fiddle)
This will actually update existing rows, and delete redundant rows, meaning you can just select from the table to get the desired results. If ofcourse you don't want to actually delete from the database you could just insert the data into a temp table and apply the same principle (Example here). In my case this solution ran a lot faster than using a recursive CTE, because at each stage of the loop the query is dealing with less data, rather than more as with the recursive cte.
WHILE EXISTS
( SELECT 1
FROM T
INNER JOIN T T2
ON T2.EMP_ID = T.EMP_ID
AND T2.DEPT_ID = T.DEPT_ID
AND T2.EMP_EFF_FROM_DATE > T.EMP_EFF_TO_DATE
AND T2.EMP_EFF_FROM_DATE <= DATEADD(DAY, 45, T.EMP_EFF_TO_DATE)
)
BEGIN
UPDATE T
SET EMP_EFF_TO_DATE = T2.EMP_EFF_TO_DATE
FROM T
INNER JOIN
( SELECT *
FROM T
) T2
ON T2.EMP_ID = T.EMP_ID
AND T2.DEPT_ID = T.DEPT_ID
AND T2.EMP_EFF_FROM_DATE > T.EMP_EFF_TO_DATE
AND T2.EMP_EFF_FROM_DATE <= DATEADD(DAY, 45, T.EMP_EFF_TO_DATE)
DELETE T
FROM T
WHERE EXISTS
( SELECT 1
FROM T T2
WHERE T2.EMP_ID = T.EMP_ID
AND T2.DEPT_ID = T.DEPT_ID
AND T2.EMP_EFF_FROM_DATE < T.EMP_EFF_FROM_DATE
AND T2.EMP_EFF_TO_DATE BETWEEN T.EMP_EFF_FROM_DATE AND T.EMP_EFF_TO_DATE
)
END;
SELECT *
FROM T
ORDER BY EMP_ID, DEPT_ID, EMP_EFF_FROM_DATE;
All of these solutions differ to your sample data in the last row which appears to be an error:
I think this row:
10 10007 2007-08-25 2007-06-29
should be:
10 10007 2007-08-25 2008-05-25

Answer 2

Assuming the next row is according to the emp_eff_from_date field (sorted), here is a way to solve it:
WITH DATA
AS (SELECT *,
Row_number()
OVER (
PARTITION BY EMP_ID
ORDER BY EMP_EFF_FROM_DATE)rn
FROM TEST)
SELECT t1.*
FROM DATA t1
INNER JOIN DATA t2
ON t1.RN = t2.RN - 1
WHERE Datediff(DAY, t1.EMP_EFF_TO_DATE, t2.EMP_EFF_FROM_DATE) <= 45
The full solution is here
Let me know if it's not exactly what you wanted.