Applying Running Total to a Result set

• Download source code - 1.8 KB

Introduction

Given the same data structures and data as in the original tip, the running total can be calculated also using a single statement.

Although the results would be the same, often using a single SQL statements gives better performance results. This is because the optimizer can more efficiently calculate the best plan to fetch the data. This optimization isn't possible over several statements since the optimizer isn't optimizing the 'general throughput' of a T-SQL block but the throughput of a single statement.

Variation 1, correlated scalar query

The first variation is to use a correlated scalar query.

-- Fetch the data using correlated scalar
SELECT  A.ProductID, 
        A.[DATE], 
        A.QTY, 
        (   SELECT SUM(B.QTY) 
            FROM
            (   SELECT ProductID, [DATE], QTY FROM GRN 
                UNION ALL
                SELECT ProductID, [DATE], - (QTY) FROM Invoice 
            ) AS B
            WHERE B.ProductId = A.ProductId
            AND   B.[DATE] <= A.[DATE]
        ) AS RuningTotal 
FROM    (    SELECT ProductID, [DATE], QTY FROM GRN 
            UNION ALL
            SELECT ProductID, [DATE], - (QTY) FROM Invoice 
        ) AS A
ORDER BY A.[Date] 

The outer body of the statement is the same as in the original tip. The difference is that for each row that is fetched frmo the two tables a scalar query is executed when the row is returned. In the scalar query the same base data is used, but the sum of QTY is calculated only for the rows for the same or previous date and for the same product.

Variation 2, GROUP BY with join

A slighly different approach is to use a GROUP BY clause and joining the data

-- Fetch the data using group by
SELECT  A.ProductID, 
        A.[DATE], 
        A.QTY, 
        SUM(B.QTY) AS RuningTotal 
FROM    (   SELECT ProductID, [DATE], QTY FROM GRN 
            UNION ALL
            SELECT ProductID, [DATE], - (QTY) FROM Invoice 
        ) AS A,
        (    SELECT ProductID, [DATE], QTY FROM GRN 
             UNION ALL
             SELECT ProductID, [DATE], - (QTY) FROM Invoice 
        ) AS B
WHERE B.ProductId = A.ProductId
AND   B.[DATE] <= A.[DATE]
GROUP BY A.ProductID, 
         A.[DATE], 
         A.QTY
ORDER BY A.[Date]

Basically the above is very much the same as variation 1. Only this time the same data is fetched twice (note, logically) and then each row from result set named A is joined to rows from results set B. The join condition is the same as previously: Same or earlier date for the same product.

What variation to use

There are also lots of other variations that could be written so what to use? Well, there's no one and only correct answer, it depends. It depends on factors like:

• Your data, this affects quite a lot. How much rows you have, what kind of statistical distribution there is for the data used in conditions and so on
• The plan, using actual (or predicted) data you should always check which variation yields the best results in terms of logical reads, CPU usage etc
• Maintainability, is this going to be hard to understand or to maintain later. If you're creating a massive piece of SQL it may be hard to understand by other people or even by yourself later. Sometimes it's better to break the problem into smaller, maintainable pieces and sometimes to squeeze all the power from the database.

History

• 14^th May, 2012: Alternative created