Home/ Questions/Q 6139087
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-05-23T17:56:45+00:00

I have a scalar-valued function called DATEONLY that returns DATEADD(DD,0, DATEDIFF(DD,0, @DATETIME)), just like

  • 0

I have a scalar-valued function called DATEONLY that returns DATEADD(DD,0, DATEDIFF(DD,0, @DATETIME)), just like this:

CREATE FUNCTION [DBO].[DATEONLY] ( @DATETIME DATETIME )
RETURNS DATETIME
BEGIN
    RETURN DATEADD(DD, 0, DATEDIFF(DD, 0, @DATETIME))
END

When I select a table using my function, the SQL Server Profiler counts a high RowCounts number than if I used directly DATEADD(DD,0, DATEDIFF(DD,0, @DATETIME)).

In my Dropxbox’s Public Folder you can find a script.sql that can reproduce what I am talking about and you can also find a Trace.trc from my SQL Server Profiler.

script.sql: https://www.dropbox.com/s/gwbh54jqas7fhhc/script.sql

trace.trc: https://www.dropbox.com/s/gwbh54jqas7fhhc/Trace.trc

Just to simplify, look the RowCounts below.

SELECT DATEADD(DD,0, DATEDIFF(DD,0, INCOMING)) AS DATA, COUNT(*) AS SOULS
FROM HELL
GROUP BY DATEADD(DD,0, DATEDIFF(DD,0, INCOMING))

RowCounts = 6

SELECT DBO.DATEONLY(INCOMING) AS DATA, COUNT(*) AS SOULS
FROM HELL
GROUP BY DBO.DATEONLY(INCOMING)

RowCounts = 32

In my real scenario, those 32 rows turns into millions RowCounts. If they are THE SAME THING, what’s going on?! How can i optimize that to prevent changing my whole application?

Thanks a lot!

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    Scalar-valued user-defined function execution is not very efficient in sql server – it essentially does a separate execution call for each invocation, which is every row in the table. Adam Machanic has a good post on this subject which describes scalar udf execution and how inline table-valued function execution can be much faster.

    One can rewrite your query to take advantage of the logic within a tvf instead, which the optimizer executes using the same query plan as your original expanded query and shows the same RowCounts=5 during execution.

    CREATE FUNCTION [DBO].[DATEONLY2] ( @DATETIME DATETIME ) RETURNS TABLE
AS RETURN SELECT DATEADD(DD, 0, DATEDIFF(DD, 0, @DATETIME)) data

select data, count(*) as souls
from
(SELECT (select data from dbo.dateonly2(incoming)) data
FROM HELL) t
GROUP BY data
    • 0