Question

The heading of this question is probably poorly worded as I am finding it difficult to explain concisely what I want, other than to provide some demo data.

I have a query which returns the following data from a sql table:

ID    Job    User    Amount

1     101    Bob     100
2     101    Pete    500
3     102    Bob     400
4     102    Pete    200
5     101    Pete    850
6     102    Bob     650

What I want is the query to also return an additional field called (Difference), which contains the difference between the Amount in consecutive entries for the same User and Job. Hence the data I would like returned would be as follows:

ID    Job    User    Amount   Diff

1     101    Bob      100     100
2     101    Pete     500     500
3     102    Bob      400     400
4     102    Pete     200     200
5     101    Pete     850     350
6     102    Bob      650     250

In the first four rows, the Diff is the same as the Amount because each is the first entry per User per Job (hence the Difference is calculated with reference to a starting Amount of nil in effect).

The last two lines contain information for a User and Job combination that have appeared in the table previously, and hence Diff is calculated as follows:

Job 101    User Pete    850 - 500 = 350
Job 102    User Bob     650 - 400 = 250

I’ve never had to compare data from rows like this in a SQL query before so don’t really know where to start. Any help would be much appreciated.

Added

Please note the Amount is not a running total. It is a subjective assessment made periodically of the value of a User’s input in each particular job. It is possible that the Amount could in fact go down from one assessment to the next. What I want is a query that returns the difference between successive assessments ‘Amounts’.

Alternative Explanation

I’m looking to return a history trail of movements in the Amount assessed. So another example, looking at a single Job and User is as follows:

Job    User    Amount    Movement

101    Bob      100       100
101    Bob      500       400
101    Bob      400      (100)
101    Bob    1,000       600

However, as per the original example, this information will need to be extracted from a table which contains many Jobs and Users all intermingled.

Answer 1

For SQL Server 2012, try this

This assumes that ID=5 value is wrong in your example

For “previous value” per pair

DECLARE @t TABLE (ID int, Job int, Username varchar(10), Amount int);
INSERT @t
VALUES
    (1, 101, 'Bob', 100), (2, 101, 'Pete', 500), (3, 102, 'Bob', 400),
    (4, 102, 'Pete', 200), (5, 101, 'Pete', 850), (6, 102, 'Bob', 650);

SELECT
    t1.*,
    t1.Amount - ISNULL(LAG(Amount) OVER (PARTITION BY Job, Username ORDER BY ID), 0) AS DiffAmount
FROM
    @t t1
ORDER BY
   t1.ID

For “first value” per pair

SELECT
    t1.*,
    CASE
        WHEN FIRST_VALUE(t1.ID) OVER (PARTITION BY Job, Username ORDER BY ID) = t1.ID THEN t1.Amount
        ELSE t1.Amount - FIRST_VALUE(t1.Amount) OVER (PARTITION BY Job, Username ORDER BY ID)
    END AS DiffAmount
FROM
    @t t1
ORDER BY
    t1.ID