Showing History of changes from a History table - sql

Question

I have a History table, which is created by Insert and Update triggers. The History row contains the row as it was on the Insert/Update.
What I am being asked to do is show the changes for each user through time. So, below is what I have in the form of my History table, and then, I created a dummy expected results.
DECLARE #MyTable TABLE
(
id INT NOT NULL IDENTITY(1,1),
userId INT NOT NULL,
locationId INT NOT NULL,
roleId INT NOT NULL,
lastUpdateUserId INT NOT NULL,
lastUpdateDate DATETIME NOT NULL
)
INSERT INTO #MyTable
(userId, locationId, roleId, lastUpdateUserId, lastUpdateDate)
SELECT 1, 1000, 1, 7, GETDATE()+1 UNION
SELECT 2, 1100, 5, 9, GETDATE()+2 UNION
SELECT 2, 1110, 5, 6, GETDATE()+3 UNION
SELECT 1, 1100, 3, 6, GETDATE()+4 UNION
SELECT 4, 1500, 5, 8, GETDATE()+5 UNION
SELECT 7, 1000, 8, 9, GETDATE()+6 UNION
SELECT 7, 1100, 9, 9, GETDATE()+7 UNION
SELECT 1, 1000, 3, 7, GETDATE()+8 UNION
SELECT 9, 1100, 5, 2, GETDATE()+9 UNION
SELECT 9, 1100, 6, 5, GETDATE()+10
SELECT * FROM #MyTable ORDER BY Id
DECLARE #ExpectedResult TABLE
(
ChangeType CHAR(1), -- I=Insert, U=Update
UserId INT,
ChangeDate DATETIME,
ChangedByUser INT,
FieldName VARCHAR(20),
OldValue INT,
NewValue INT
)
INSERT INTO #ExpectedResult
(ChangeType, UserId, ChangeDate, ChangedByUser, FieldName, OldValue, NewValue)
SELECT 'I', 1, '2015-APR-30 09:56:28', 7, 'locationId', NULL, 1000 UNION -- Row1
SELECT 'I', 1, '2015-APR-30 09:56:28', 7, 'roleId', NULL, 1 UNION -- Row1
SELECT 'U', 1, '2015-APR-07 10:27:42', 7, 'roleId', 1, 3 UNION -- Row 2
SELECT 'U', 1, '2015-MAY-03 10:27:42', 6, 'locationId', 1000, 1100 UNION -- Row 3
SELECT 'I', 2, '2015-MAY-01 10:27:42', 9, 'roleId', NULL, 5 UNION -- Row5
SELECT 'I', 2, '2015-MAY-01 10:27:42', 9, 'locationId', NULL, 1100 -- Row5
SELECT * FROM #ExpectedResult
#MyTable has the data as it is at the moment. I am trying to transform that into #ExpectedResults. We're reporting on changes to roleId and locationId. On each change, it needs to have a separate line for each column. So, on insert, we have two lines (As we monitor changes to both fields). When one column is updated, it needs to be represented as one 'U' line. If both fields are updated in the same UPDATE statement, then that would result in two update rows in #Expected.
I started with a Cursor, but was hoping there would be a more efficient way to achieve this.

Answer 1

To get the roleID and locationID on separate rows you can use a simple UNION ALL.
And to combine the old and new values use the ROW_NUMBER() window function, like this:
;with t as(
select *,
ROW_NUMBER() OVER(partition by userid Order BY lastUpdateDate) rn
from #MyTable
),
a as (
select userId, 'locationId' as fieldname,
locationId as value, lastUpdateUserId, lastUpdateDate, rn
from t
UNION ALL
select userId, 'roleId' as fieldname,
roleId as value, lastUpdateUserId, lastUpdateDate, rn
from t
)
select CASE WHEN a2.userId IS NULL THEN 'I' ELSE 'U' END as ChangeType,
a1.userId, a1.lastUpdateDate, a1.lastUpdateUserId, a1.fieldname, a1.value as newValue, a2.value as oldvalue
FROM a a1 LEFT JOIN a a2
ON a1.userId = a2.userId and a1.fieldname = a2.fieldname
AND a1.rn = a2.rn+1
order by 2,3,5
The a1 alias in the query above contains the "new values", the a2 contains the "old values". When you use the real data you will also need to partition by the fieldname (and perhaps table name) and also to join by them
The result:
ChangeType userId lastUpdateDate lastUpdateUserId fieldname newValue oldvalue
---------- ----------- ----------------------- ---------------- ---------- ----------- -----------
I 1 2015-04-30 12:20:59.183 7 locationId 1000 NULL
I 1 2015-04-30 12:20:59.183 7 roleId 1 NULL
U 1 2015-05-03 12:20:59.183 6 locationId 1100 1000
U 1 2015-05-03 12:20:59.183 6 roleId 3 1
U 1 2015-05-07 12:20:59.183 7 locationId 1000 1100
U 1 2015-05-07 12:20:59.183 7 roleId 3 3
I 2 2015-05-01 12:20:59.183 9 locationId 1100 NULL
I 2 2015-05-01 12:20:59.183 9 roleId 5 NULL
U 2 2015-05-02 12:20:59.183 6 locationId 1110 1100
U 2 2015-05-02 12:20:59.183 6 roleId 5 5
I 4 2015-05-04 12:20:59.183 8 locationId 1500 NULL
I 4 2015-05-04 12:20:59.183 8 roleId 5 NULL
I 7 2015-05-05 12:20:59.183 9 locationId 1000 NULL
I 7 2015-05-05 12:20:59.183 9 roleId 8 NULL
U 7 2015-05-06 12:20:59.183 9 locationId 1100 1000
U 7 2015-05-06 12:20:59.183 9 roleId 9 8
I 9 2015-05-08 12:20:59.183 2 locationId 1100 NULL
I 9 2015-05-08 12:20:59.183 2 roleId 5 NULL
U 9 2015-05-09 12:20:59.183 5 locationId 1100 1100
U 9 2015-05-09 12:20:59.183 5 roleId 6 5
(20 row(s) affected)