Author: Kevin Feasel

Daniel Hutmacher explains how SET STATISTICS XML will generate execution plans for certain segments of code:

But sometimes you want to run a series of statements or procedures where you only want the execution plan for some of the statements. Here’s how:

The actual execution plan is enabled by turning on SET STATISTICS XML., not unlike enabling STATISTICS IO or TIME. And just like SET NOCOUNT, the SET statements apply to the current context, which could be a stored procedure, a session, etc. When this context ends, the setting reverts to that of the parent context.

I see code snippets with STATISTICS IO and TIME fairly regularly, but almost never see STATISTICS XML; instead, I see people (including myself) hit Ctrl-M or select the “Include Actual Execution Plan” button when generating execution plans is desirable.

James Anderson shows a case in which an update with a join clause can lead to undesirable results:

That’s right; SQL Server will just pick any child. It will not update the parent row for each child. When the MERGE statement runs into this problem, it raises an error and rolls back the UPDATE but our query will silently pick any value and move on. We don’t want to update data like this, we want our query to dictate the logic explicitly. This will require some changes to our query.

I prefer to write these types of updates with the use of a CTE. This allows us to easily highlight the SELECT in the CTE and execute it to see which rows will be updated (@variables can cause problems here though). Adding a COUNT can help to identify the problem in the previous query.

Highlight and execute the code inside the CTE below to check for duplicates. If this returns 0 rows then you are good to remove the COUNT, GROUP BY and HAVING then add the name columns before executing the whole statement.

When writing T-SQL updates with joins, it’s important to consider whether the grain changes, and if that change can make a difference in your update set.