When to use Common Table Expression (CTE)

There are two reasons I see to use cte's.

To use a calculated value in the where clause. This seems a little cleaner to me than a derived table.

Suppose there are two tables - Questions and Answers joined together by Questions.ID = Answers.Question_Id (and quiz id)

WITH CTE AS
(
    Select Question_Text,
           (SELECT Count(*) FROM Answers A WHERE A.Question_ID = Q.ID) AS Number_Of_Answers
    FROM Questions Q
)
SELECT * FROM CTE
WHERE Number_Of_Answers > 0

Here's another example where I want to get a list of questions and answers. I want the Answers to be grouped with the questions in the results.

WITH cte AS
(
    SELECT [Quiz_ID] 
      ,[ID] AS Question_Id
      ,null AS Answer_Id
          ,[Question_Text]
          ,null AS Answer
          ,1 AS Is_Question
    FROM [Questions]

    UNION ALL

    SELECT Q.[Quiz_ID]
      ,[Question_ID]
      ,A.[ID] AS  Answer_Id
      ,Q.Question_Text
          ,[Answer]
          ,0 AS Is_Question
        FROM [Answers] A INNER JOIN [Questions] Q ON Q.Quiz_ID = A.Quiz_ID AND Q.Id = A.Question_Id
)
SELECT 
    Quiz_Id,
    Question_Id,
    Is_Question,
    (CASE WHEN Answer IS NULL THEN Question_Text ELSE Answer END) as Name
FROM cte    
GROUP BY Quiz_Id, Question_Id, Answer_id, Question_Text, Answer, Is_Question 
order by Quiz_Id, Question_Id, Is_Question Desc, Name

One of the scenarios I found useful to use CTE is when you want to get DISTINCT rows of data based on one or more columns but return all columns in the table. With a standard query you might first have to dump the distinct values into a temp table and then try to join them back to the original table to retrieve the rest of the columns or you might write an extremely complex partition query that can return the results in one run but in most likelihood, it will be unreadable and cause performance issue.

But by using CTE (as answered by Tim Schmelter on Select the first instance of a record)

WITH CTE AS(
    SELECT myTable.*
    , RN = ROW_NUMBER()OVER(PARTITION BY patientID ORDER BY ID)
    FROM myTable 
)
SELECT * FROM CTE
WHERE RN = 1

As you can see, this is much easier to read and maintain. And in comparison to other queries, is much better at performance.

One example, if you need to reference/join the same data set multiple times you can do so by defining a CTE. Therefore, it can be a form of code re-use.

An example of self referencing is recursion: Recursive Queries Using CTE

For exciting Microsoft definitions Taken from Books Online:

A CTE can be used to:

• Create a recursive query. For more information, see Recursive Queries Using Common Table Expressions.

• Substitute for a view when the general use of a view is not required; that is, you do not have to store the definition in metadata.

• Enable grouping by a column that is derived from a scalar subselect, or a function that is either not deterministic or has external access.

• Reference the resulting table multiple times in the same statement.

I use them to break up complex queries, especially complex joins and sub-queries. I find I'm using them more and more as 'pseudo-views' to help me get my head around the intent of the query.

My only complaint about them is they cannot be re-used. For example, I may have a stored proc with two update statements that could use the same CTE. But the 'scope' of the CTE is the first query only.

Trouble is, 'simple examples' probably don't really need CTE's!

Still, very handy.