MyDoggieJessie (3/7/2014)
To add to SQLCJ's post, you will need to include the ORDER BY date_value at the end to get the OP's expected result - and some ugly string manipulation for the dateSELECT CONVERT(CHAR(4), GETDATE(), 120) + ' ' + CONVERT(CHAR(4), GETDATE(), 100) ,
monthly_total ,
( SELECT SUM(monthly_total)
FROM #sales s
WHERE date_value <= tst.date_value
) AS monthly_total
FROM ( SELECT * ,
ROW_NUMBER() OVER ( ORDER BY date_value ) tst
FROM #sales
) tst
ORDER BY date_value A
Ah, be carefull now. That contains an (X2+X)/2+X Triangular join. In this case, it generates 324 internal rows for the 24 original rows. While it "runs fast" for such a small quantity of rows, these things can be server killers. With just 10,000 original rows, SQL Server will generate 50,015,000 internal rows and all of the logical reads to go with it. A cursor and WHILE loop or a Recursive CTE would actually run much faster.
Please see the following article about "Triangular Joins" for more information.
http://www.sqlservercentral.com/articles/T-SQL/61539/
OR... if you follow the rules, there's a bit of undocumented code known at the "Quirky Update". If you don't follow the rules, it can bite ya. It will, however, do a million row running or grouped running total in just seconds and is faster than even the new methods in SQL Server 2012. Plese see the following article for that.
http://www.sqlservercentral.com/articles/T-SQL/68467/
--Jeff Moden
Change is inevitable... Change for the better is not.