Query to reflect actual significant change in data - sql

Question

Given a table with employee statuses and effective dates, how can I retrieve just the data that reflects a change in status?
For example, given the following structure:
DECLARE #STATUSES TABLE(
EMPLOYEE_ID INT NOT NULL,
EFFECTIVE_DATE DATE NOT NULL,
STATUS_CODE CHAR(1) NOT NULL
)
INSERT #STATUSES VALUES (1, '2012-01-01', 'A')
INSERT #STATUSES VALUES (1, '2012-02-28', 'A')
INSERT #STATUSES VALUES (1, '2012-03-01', 'T')
INSERT #STATUSES VALUES (2, '2012-01-01', 'A')
INSERT #STATUSES VALUES (2, '2012-02-14', 'A')
INSERT #STATUSES VALUES (2, '2012-03-10', 'A')
INSERT #STATUSES VALUES (3, '2012-02-01', 'A')
INSERT #STATUSES VALUES (3, '2012-03-17', 'A')
INSERT #STATUSES VALUES (3, '2012-03-18', 'T')
INSERT #STATUSES VALUES (3, '2012-04-01', 'A')
INSERT #STATUSES VALUES (4, '2012-03-01', 'A')
What query can be used to result in the following?
EMPLOYEE_ID EFFECTIVE_DATE STATUS_CODE
1 2012-01-01 A
1 2012-03-01 T
2 2012-01-01 A
3 2012-02-01 A
3 2012-03-18 T
3 2012-04-01 A
4 2012-03-01 A
In other words, I want to leave out those records that have the same employee id and status code as the one before it, if one exists with an earlier effective date. Notice that employee 1 is listed only twice because there were only two actual changes in status--the one on 2012-02-28 is inconsequential since the status didn't change from the earlier date. Also notice that employee 2 is listed just once since his status never changed despite there being three records. Only the earliest date is shown for each change.

Answer 1

With some further experimenting, it looks like this will do what I want.
;WITH cte
AS (SELECT ROW_NUMBER() OVER (PARTITION BY EMPLOYEE_ID ORDER BY EFFECTIVE_DATE) AS rownum
,EMPLOYEE_ID
,EFFECTIVE_DATE
,STATUS_CODE
FROM #STATUSES)
SELECT t2.EMPLOYEE_ID
,t2.EFFECTIVE_DATE
,t2.STATUS_CODE
FROM cte t2
LEFT JOIN cte t1
ON t2.EMPLOYEE_ID = t1.EMPLOYEE_ID
AND t2.STATUS_CODE = t1.STATUS_CODE
AND t2.rownum = t1.rownum + 1
WHERE t1.EMPLOYEE_ID IS NULL

Answer 2

You could use a CURSOR
You'd need two sets of variables: #PreviousRecord and #CurrentRecord
Declare the cursor for table sorted by employeeid and date
Fetch the first record from the cursor into the #PreviousRecord variables - depending on your requirement register this as a significant change or not (write the record to a temp table)
Then set up a loop that:
Fetches the next record into the #CurrentRecord variables
Compares it with the previous record and if it matches your requirement for a significant change write it to the temp table
Move the #CurrentRecord values into the #PreviousRecord variables
I'd be interested to know if the CTE method was more efficient

Answer 3

SELECT
EMPLOYEE_ID, MIN(EFFECTIVE_DATE) AS EFFECTIVE_DATE, STATUS_CODE
FROM
(
SELECT
T1.EMPLOYEE_ID, T1.EFFECTIVE_DATE, T1.STATUS_CODE,
MAX(T2.EFFECTIVE_DATE) AS MOST_RECENT_PREVIOUS_STATUS_DATE
FROM
#STATUSES T1
LEFT JOIN
#STATUSES T2
ON
T1.EMPLOYEE_ID = T2.EMPLOYEE_ID
AND
T1.EFFECTIVE_DATE > T2.EFFECTIVE_DATE
AND
T1.STATUS_CODE <> T2.STATUS_CODE
GROUP BY
T1.EMPLOYEE_ID, T1.EFFECTIVE_DATE, T1.STATUS_CODE
) SubQuery
GROUP BY
EMPLOYEE_ID, STATUS_CODE, MOST_RECENT_PREVIOUS_STATUS_DATE