Question

We are using audit tables for each operational table, which stores the previous value of its operational equivalent plus change date, change type (UPDATE or DELETE) and its own auto incremental Primary Key.

So, for a table Users with columns UserID, Name, Email there would be a table xUsers with columns ID, OpererationType, OperationDate, UserID, Name, Email.

See that the xTable contains every column that its ‘parent’ does with 3 extra fields. This pattern is repeated for all tables used by our system.

table Users:

UserID           int    
Name             nvarchar
Email            nvarchar

table xUsers:

    xUserID          int    
    OpererationType  int
    OperationDate    datetime
    UserID           int    
    Name             nvarchar
    Email            nvarchar

Now, my question:
If I have a certain UserID, for which there is 2 entries in the xUsers table when the email was changed twice,

how would I construct a query that identifies which columns (can be more than 1) differ between the two rows in the audit table?

Answer 1

If I’m understanding this correctly, you’d like to create a query passing in the UserID as a parameter, which I’ll call @UserID for the following example.

This query will select all rows from xUsers joined onto itself where there is a difference in a non-UserID column, using a series of case statements (one per column) to pull out specifically which columns differ.

SELECT *
    , CASE 
        WHEN a.OperationType <> b.OperationType
        THEN 1
        ELSE 0
        END AS OperationTypeDiffers
    , CASE
        WHEN a.OperationDate <> b.OperationDate
        THEN 1
        ELSE 0
        END AS OperationDateDiffers
FROM xUsers a
JOIN xUsers b
    ON a.xUserID < b.xUserID
    AND a.UserID = b.UserID
    AND (a.OperationType <> b.OperationType
        OR a.OperationDate <> b.OperationDate) -- etc.
WHERE a.UserID = @UserID