Why would a stored procedure return a 0 and the selected id? - sql

Question

I have a stored procedure like this:
CREATE PROCEDURE [dbo].[create_myNewId]
(#parentId BIGINT)
AS
BEGIN
SET NOCOUNT ON;
INSERT INTO [Mapping] (ParentId)
VALUES (#parentId)
SELECT SCOPE_IDENTITY();
END
This, when run on its own, returns the new id that has been assigned to the new row that's inserted with the parent id. However, when I do something like this:
DECLARE #NewId int
EXEC #NewId = create_myNewId #parentId = 33333
SELECT #NewId
When running this, the output window shows the result of the stored procedure, which returns an Id but #NewId always is 0. I fixed this by changing the stored procedure to use RETURN SCOPE_IDENTITY() but I was wondering why SELECT didn't work in this case?
I have my suspicions that it's something around the 0 being the success status being returned first from the stored procedure rather than the result, but was curious why this doesn't then happen when called directly from the client.

Answer 1

No! Write the procedure the right way:
CREATE PROCEDURE [dbo].[create_myNewId] (
#parentId bigint,
#outId bigint OUTPUT
) AS
BEGIN
SET NOCOUNT ON;
DECLARE #ids TABLE (id bigint);
INSERT INTO [Mapping](ParentId)
OUTPUT id INTO #ids
VALUES (#parentId);
SELECT #outId = id
FROM #ids;
END;
Then call this as:
DECLARE #NewId int;
EXEC create_myNewId #parentId = 33333, #NewId OUTPUT;
SELECT #NewId;
The OUTPUT clause is the recommend way to get results from a data-modification clause. The older methods using the *_IDENTITY() functions should be obsoleted.
Stored procedures do return values. These are integers that are designed to return status information. Other information should be returned via OUTPUT parameters.

Answer 2

Microsoft's design intent for stored procedures is that they always return an int to describe how successful the process undertaken by the procedure was. It's not intended to return a result data, and you're free to define the bits you want to return to describe succes, partial success etc. You could abuse it to return an integer result data (count query for example) if you wanted, but it's not the design intention
Executing a select query within a stored procedure creates a result set you can read on your client if the sproc is the kind that is intended to return data

Answer 3

My suggestion is to use an OUTPUT parameter. Not only will it be 'easier' to use when calling the stored procedure, it will also be clearer to the person calling the stored procedure.
CREATE PROCEDURE [dbo].[create_myNewId]
(#parentId BIGINT,
#myNewId BIGINT OUTPUT)
AS
BEGIN
SET NOCOUNT ON;
INSERT INTO [Mapping] ([ParentId])
VALUES (#parentId);
SET #myNewId = SCOPE_IDENTITY();
END;
GO
You would then call your stored procedure like this:
DECLARE #myNewId BIGINT;
EXECUTE [dbo].[create_myNewId] #parentId = 0, -- bigint
#myNewId = #myNewId OUTPUT; -- bigint
SELECT [This was just inserted] = #myNewId;

Answer 4

For anyone who has 0 as return value from a stored procedure, check if the stored procedure executes from the right database and only one procedure exists within the given context. Output parameters wouldn't be of any use if you ever plan to access the DB with ORM and the procedure returns an object's property.