Click here to monitor SSC
SQLServerCentral is supported by Red Gate Software Ltd.
 
Log in  ::  Register  ::  Not logged in
 
 
 

SQLStudies

My name is Kenneth Fisher and I am Senior DBA for a large (multi-national) insurance company. I have been working with databases for over 20 years starting with Clarion and Foxpro. I’ve been working with SQL Server for 12 years but have only really started “studying” the subject for the last 3. I don’t have any real "specialities" but I enjoy trouble shooting and teaching. Thus far I’ve earned by MCITP Database Administrator 2008, MCTS Database Administrator 2005, and MCTS Database Developer 2008. I’m currently studying for my MCITP Database Developer 2008 and should start in on the 2012 exams next year. My blog is at www.sqlstudies.com.

Using the OVER clause

Over the last few years I’ve learned quite a bit about different techniques in SQL Server. This particular one has been available since SQL 2005 and looks like it could be really handy.

Let’s say I want to calculate the total sales for by month and at the same the percentage of sales that month represents for the year.

Without the OVER clause I would have to do something with a CTE or a Subquery. For example:

WITH YearTotals AS (
	SELECT YEAR(OrderDate) AS OrderYear,
		SUM(TotalDue) AS YearTotal
	FROM Sales.SalesOrderHeader
	GROUP BY YEAR(OrderDate)
	)
SELECT YEAR(OrderDate) AS OrderYear, MONTH(OrderDate)AS OrderDate,
	SUM(TotalDue) AS MonthTotal,
	YearTotals.YearTotal,
	SUM(TotalDue)/YearTotals.YearTotal*100 AS MonthlyPercentage
FROM Sales.SalesOrderHeader
JOIN YearTotals
	ON YearTotals.OrderYear = YEAR(SalesOrderHeader.OrderDate)
GROUP BY YEAR(OrderDate), MONTH(OrderDate), YearTotals.YearTotal
ORDER BY YEAR(OrderDate), MONTH(OrderDate)

Using the OVER clause I’m able to specify the criteria for each aggregate like so.

SELECT DISTINCT YEAR(OrderDate) AS OrderYear, MONTH(OrderDate)AS OrderDate,
	SUM(TotalDue) OVER (PARTITION BY YEAR(OrderDate), MONTH(OrderDate)) AS MonthTotal,
	SUM(TotalDue) OVER (PARTITION BY YEAR(OrderDate)) AS YearTotal,
	SUM(TotalDue) OVER (PARTITION BY YEAR(OrderDate), MONTH(OrderDate)) /
		SUM(TotalDue) OVER (PARTITION BY YEAR(OrderDate))*100 AS MonthlyPercentage
FROM Sales.SalesOrderHeader
ORDER BY YEAR(OrderDate), MONTH(OrderDate)

The OVER version looks quite a bit simpler doesn’t it? And I’ve always thought simpler usually meant faster and more efficient. Interestingly enough however, when I looked at performance the CTE version was significantly more efficient. The OVER version used almost 200 times as many reads and took almost 7 times as long to run. Now this was just one particular example, but still something to be watched. So, while I can think of several cases where using the OVER clause would have been a quick way to add a total to a query, I think I’m still going to have to carefully test its performance before putting it into production.

Note: My initial version of the OVER query used a GROUP BY to calculate the MonthTotal and then the OVER to calculate the YearTotal. Unfortunately I kept getting an error. I’m not entirely sure if I’m making a mistake or if OVER and GROUP BY are just incompatible. I’ll continue to research and if I find out one way or the other I’ll add it to the post.


Filed under: Microsoft SQL Server, SQLServerPedia Syndication, T-SQL Tagged: code language, language sql, microsoft sql server, sql statements, T-SQL

Comments

Leave a comment on the original post [sqlstudies.com, opens in a new window]

Loading comments...