Question

I am using the Spring Framework’s StoredProcedure (I am extending it, of course) to get a result set and an output parameter (@totalRowsReturned) which is an Integer. The problem is that when the resultset being returned is supposed to be an empty list, I am getting a NullPointerException when I try to retrieve the output parameter (totalRows, which naively I would expect it to be zero).

I would like to mention that the code works fine when the result set being found is not empty.

My questions are:

1. Why isn’t @totalRowsReturned being set to zero in this case? (Or in case it is, why can’t I retrieve it through the Java code?)

2. How can I make this code (Java code + T-SQL code) work in such a way that @totalRowsReturned will be set to zero when required, and I could retrieve it through the Java code?

Dao:

List<Book> books = null;
int totalRows = 0;

Map<String, Object> results = storedProcedure.execute(parameters);
books = (List<Book>) results.get("rs");
totalRows = (Integer) results.get("totalRowsReturned"); // NullPointerException on this line if total rows are supposed to be zero!!

T-SQL stored procedure:

CREATE PROCEDURE Find_Books

    @authorName Varchar(250),   
    @totalRowsReturned INTEGER OUTPUT

AS

BEGIN

    SET NOCOUNT ON;
    DECLARE @SelectQuery NVARCHAR(2000)

    SET @SelectQuery = 'SELECT @totalRows=COUNT(*) OVER() FROM book b WHERE b.author_name = @authorName'

    Execute sp_Executesql @SelectQuery, N'@authorName VARCHAR(250), @totalRows int OUTPUT', @authorName, @totalRows=@totalRowsReturned OUTPUT

-- Select resultset goes here...

END

UPDATE:

Actually, my stored procedure looks more like this (the change is the additional @first_id = b.book_id in the SELECT):

CREATE PROCEDURE Find_Books

    @authorName Varchar(250),   
    @totalRowsReturned INTEGER OUTPUT

AS

BEGIN

    SET NOCOUNT ON;
    DECLARE @SelectQuery NVARCHAR(2000)

    DECLARE @first_id int
    DECLARE @first_id_local_returned int

    SET @SelectQuery = 'SELECT @first_id = b.book_id, @totalRows=COUNT(*) OVER() FROM book b WHERE b.author_name = @authorName'

    Execute sp_Executesql @SelectQuery, N'@authorName VARCHAR(250), @first_id int OUTPUT, @totalRows int OUTPUT', @authorName, @first_id=@first_id_local_returned OUTPUT, @totalRows=@totalRowsReturned OUTPUT

    -- Select resultset goes here... I am using the value of @first_id_local_returned in this SELECT...

END

The problem is, that when there are no rows returned from the SELECT, b.book_id is not defined, so I get an org.springframework.dao.TransientDataAccessResourceException ... Column 'book.book_id' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.

So it seems like if I keep the OVER(), then @totalRows=COUNT(*) OVER() fails when there are zero rows returned, and if I remove the OVER(), then @first_id = b.book_id fails.

Any idea how I overcome this?

Answer 1

COUNT(*) OVER () is not the correct thing to use here. Just use COUNT(*)

COUNT(*) OVER () returns a result set with as many rows as the COUNT. e.g if the result is 3 the result set will be

3
3
3

The effect of your query is then to repeatedly re-assign the value 3 to the @totalRows variable as many times as there are rows which is completely pointless.

Conversely if COUNT(*) = 0 then the COUNT(*) OVER () result set is empty so your variable is never assigned to at all.

COUNT(*) will always give you a single row scalar resultset here that you can assign to the variable and will have a more efficient execution plan without unnecessary common subexpression spools too.

Edit

In response to your question in the comments. This does the same thing as your linked article. It can use a narrower index to find the (say) 10,000th Employee then joins onto Department only for the 1 subsequent page of records. This paging method only works correctly because each employee has exactly one department.

WITH E1(RN, EmployeeID)
     AS (SELECT ROW_NUMBER() OVER (ORDER BY EmployeeID),
                EmployeeID
         FROM   Employees)
SELECT TOP (@maximumRows) e.*,
                          d.Name AS DepartmentName
FROM   Employees e
       INNER JOIN Departments d
         ON e.DepartmentID = d.DepartmentID
WHERE  EmployeeID >= (SELECT EmployeeID
                      FROM   E1
                      WHERE  RN = @startRowIndex)
ORDER  BY e.EmployeeID