Comments on: Retrieve the top X random rows from a query

By: Jeremiah Peschka

Jeremiah Peschka — Tue, 27 Jan 2009 12:45:33 +0000

It actually turns out that the execution plans are almost exactly the same (in this case). Using TOP (@Count) is slightly more efficient within this scenario in AdventureWorks (the plan cost for my query is 0.412827 and for TOP it is 0.412769). The difference is made up by some additional operators and a filter that I’m using to replace the functionality of TOP via my use of ROW_NUMBER and the WHERE clause.

Which brings up the point that we’re always supposed to be using the parens with TOP and I just found that out this morning while reading Books Online.

By: Rick Kierner

Rick Kierner — Tue, 27 Jan 2009 12:14:21 +0000

I just found this:
DECLARE @Count int
SELECT @Count = 5

SELECT TOP (@Count),
c.ContactId,
c.FirstName,
c.LastName,
e.EmployeeId,
a.AddressId,
a.AddressLine1,
a.AddressLine2,
a.City,
a.PostalCode,
at.Name AS AddressType
FROM HumanResources.Employee AS e
INNER JOIN Person.Contact AS c
ON e.ContactId = c.ContactId
INNER JOIN HumanResources.EmployeeAddreess AS ea
ON e.EmployeeId = ea.EmployeeId
INNER JOIN Person.Address AS a
ON ea.AddressId = a.AddressId
INNER JOIN Person.AddressType AS at
ON a.AddressTypeId = at.AddressTypeId

Is this better/worse/same?

By: Denis Gobo

Denis Gobo — Mon, 26 Jan 2009 21:29:24 +0000

you can also use SET ROWCOUNT n, still works in 2008 but will be deprecated in the next version I believe, make sure that it is set to 0 after the query

declare @id int
select @id = 5

set rowcount @id

select * from sysobjects
order by newid()

set rowcount 0

By: Rick Kierner

Rick Kierner — Mon, 26 Jan 2009 13:26:29 +0000

And thanks for your help.