Monday, June 23, 2008

Select top n random rows from a table in SQL SERVER...

There are many ways that you can use randomly selected rows; they're especially effective when you want to add dynamism to a site. For instance, you could randomly select a product to present as Today's Featured Product, or QA could generate a random call list to gauge customer satisfaction levels.

The snag is that SQL doesn't permit the selection of random rows. The good news is that there's a simple trick to getting this functionality to work in SQL.

The solution is based on the uniqueidentifierdata type. Unique identifiers, which are also called Guaranteed Unique Identifiers (GUIDs), look something like this:


SQL Server uses GUIDs in many contexts, perhaps most notably in replication. You can use them when normal incrementing identity columns won't provide a sufficient range of keys. To do this, you create a column of type uniqueidentifierwhose default value is NewID(), like this:

PK uniqueidentifier NOT NULL DEFAULT NewID(),
AnotherColumn varchar(50) NOT NULL

. . .

This function is just the ticket to solve our random rows problem. We can simply call NewID() as a virtual column in our query, like this:

select top 4 Column_Name,NEWID() AS RANDOM from TABLE order by RANDOM

No comments: