incrementally updating SQL rows - sql

Question

I'm currently dealing with two tables.
One table contains a set of columns like ID, NAME, AGE, TEAM, SCHOOL, and so forth in a table called PRIMARY_TABLE
And I also have an audit table called SECONDARY_TABLE that records updates in the aforementioned values over time.
I have ATTRIBUTE, PREV_VALUES and RECORD_ID columns in this table. It has the following attributes:
the RECORD_ID column corresponds to the ID column of PRIMARY_TABLE
the ATTRIBUTE column will store the column of the PRIMARY_TABLE that is being altered.
For example, if I have
132 NIKO 18 LANCERS JESUIT
143 KEENAN 25 RAIDERS ROCKLAND
in my first table and
132 'AGE' 22
132 'NAME' STEVAN
in my second,
then I want a combined table that has
132 NIKO 18 LANCERS JESUIT
132 NIKO 22 LANCERS JESUIT
132 STEVAN 22 LANCERS JESUIT
143 KEENAN 25 RAIDERS ROCKLAND .
The issue I have a hard time getting around is preserving the values in the unaffected rows. It seems like any idea I have for joining the two tables together won't work for this reason.
Any thoughts? I think the only solution is to create a stored procedure for this. If you need clarification, let me know as well.
EDIT
One more thing...
Here's another thing. The audit table also has a "time_of_change" column. If multiple rows have the same time of change for an ID, then instead of having multiple rows in our resulting table, there should be only one more row.
For example, if our audit table had
132 'AGE' 22 1:00
132 'NAME' STEVAN 1:00
Then instead of having
132 STEVAN 18 LANCERS JESUIT
132 NIKO 22 LANCERS JESUIT
added, there should only be one added row of
132 STEVAN 22 LANCERS JESUIT.
I can't think of any possible way to do this either.

Answer 1

UPDATE2 If you were to have a column with a datetime of an update in the secondary_table (lets call it updated_at) then you can order the resultset appropriately.
SELECT id, name, age, team, school, GETDATE() updated_at
FROM primary_table
UNION ALL
SELECT p.id,
CASE WHEN s.attribute = 'NAME'
THEN s.prev_values ELSE p.name END name,
CASE WHEN s.attribute = 'AGE'
THEN s.prev_values ELSE p.age END age,
CASE WHEN s.attribute = 'TEAM'
THEN s.prev_values ELSE p.team END team,
CASE WHEN s.attribute = 'SCHOOL'
THEN s.prev_values ELSE p.school END school,
updated_at
FROM primary_table p JOIN secondary_table s
ON p.ID = s.record_id
ORDER BY id, updated_at DESC
Here is SQLFiddle demo
UPDATE1 A version with one UNION and conditional output with CASE
SELECT *
FROM primary_table
UNION ALL
SELECT p.id,
CASE WHEN s.attribute = 'NAME'
THEN s.prev_values ELSE p.name END name,
CASE WHEN s.attribute = 'AGE'
THEN s.prev_values ELSE p.age END age,
CASE WHEN s.attribute = 'TEAM'
THEN s.prev_values ELSE p.team END team,
CASE WHEN s.attribute = 'SCHOOL'
THEN s.prev_values ELSE p.school END school
FROM primary_table p JOIN secondary_table s
ON p.ID = s.record_id
ORDER BY id
Here is SQLFiddle demo
Original version with UNIONs
SELECT *
FROM primary_table
UNION ALL
SELECT p.id, s.prev_values, p.age, p.team, p.school
FROM primary_table p JOIN secondary_table s
ON p.ID = s.record_id
AND s.attribute = 'NAME'
UNION ALL
SELECT p.id, p.name, s.prev_values, p.team, p.school
FROM primary_table p JOIN secondary_table s
ON p.ID = s.record_id
AND s.attribute = 'AGE'
ORDER BY id
Output:
| ID | NAME | AGE | TEAM | SCHOOL |
-------------------------------------------
| 132 | NIKO | 18 | LANCERS | JESUIT |
| 132 | STEVAN | 18 | LANCERS | JESUIT |
| 132 | NIKO | 22 | LANCERS | JESUIT |
| 143 | KEENAN | 25 | RAIDERS | ROCKLAND |
Here is SQLFiddle demo

Answer 2

This is not perfect but try this one:
SELECT * FROM Primary_Table
UNION ALL
SELECT * FROM
(SELECT A.ID, A.Name, B.GetAge as Age, A.School
FROM Primary_Table A INNER JOIN (SELECT Record_id, CONVERT(int,Value) as GetAge FROM Secondary_Table
WHERE Attrib='AGE' ) B
ON A.ID = B.Record_id) UAge
UNION ALL
SELECT * FROM
(SELECT A.ID, B.GetName as Name, A.Age, A.School
FROM Primary_Table A INNER JOIN (SELECT Record_id, Value as GetName FROM Secondary_Table
WHERE Attrib='NAME' ) B
ON A.ID = B.Record_id) UName
ORDER BY ID

Answer 3

Here's some script to get you started. You have a big job ahead of you. Things are made quite a bit more difficult if you don't have the current values in the audit table (which it seems like, so I built my script with that assumption).
WITH Curr AS (
SELECT
Record_ID = T.ID,
V.*,
Time_Of_Change = Convert(datetime, 0)
FROM
dbo.Team T
CROSS APPLY (VALUES
('Name', Convert(varchar(20), T.Name)),
('Age', Convert(varchar(20), T.Age)),
('Team', Convert(varchar(20), T.Team)),
('School', Convert(varchar(20), T.School))
) V (Attribute, Prev_Values)
),
Data AS (
SELECT
H.Record_ID,
H.Time_Of_Change,
C.Attribute,
A.Prev_Values
FROM
(
SELECT DISTINCT Record_ID, Time_Of_Change
FROM dbo.Audit
WHERE TableName = 'Team'
UNION ALL
SELECT DISTINCT ID, GetDate()
FROM dbo.Team
) H
CROSS JOIN (VALUES
('Name'), ('Age'), ('Team'), ('School')
) C (Attribute)
CROSS APPLY (
SELECT TOP 1 *
FROM (
SELECT Record_ID, Attribute, Prev_Values, Time_Of_Change
FROM dbo.Audit
WHERE TableName = 'Team'
UNION ALL
SELECT *
FROM Curr
) A
WHERE
H.Record_ID = A.Record_ID
AND H.Time_Of_Change >= A.Time_Of_Change
AND C.Attribute = A.Attribute
ORDER BY
A.Time_Of_Change DESC
) A
)
SELECT *
FROM
Data
PIVOT (Max(Prev_Values) FOR Attribute IN (Name, Age, Team, School)) P
;
See a Live Demo at SQL Fiddle
You can very easily create a view for each table you want to reconstruct the history of. Build a stored procedure that queries the INFORMATION_SCHEMA.COLUMNS view and builds something similar to what I've given you. Run the SP once after any table change, and it will update the views.
I noticed that my script is not quite right--it's showing 2 rows instead of 3 for the edits. Plus, there's no valid time for the rows for which there are no audit rows. But that part makes sense. In any case, you have a big job ahead of you...