Question

Let’s say I have a table

my_table(id int identity(1,1) not null primary key, data varchar(100))

I want to write a procedure that inserts a new row into that table and returns id.
I tried

DECLARE @new_id INT;    
SELECT @new_id = id FROM
(
INSERT INTO my_table(data) OUTPUT inserted.id VALUES ('test') 
) as NewVal(id)

That code doesn’t work (I got “A nested INSERT, UPDATE, DELETE, or MERGE statement is not allowed in a SELECT statement that is not the immediate source of rows for an INSERT statement.”). However, if I use a table variable, I can do

DECLARE @new_id INT;    
DECLARE @tmp_table TABLE(int id);
INSERT INTO @tmp_table
SELECT id FROM
(
INSERT INTO my_table(data) OUTPUT inserted.id VALUES ('test')  
) as NewVal(id);
// OR 
INSERT INTO my_table(data) OUTPUT inserted.id INTO @tmp_table VALUES ('test') ;
SELECT @new_id = id FROM @tmp_table;

Is it possible to achieve the same functionality without using table variable ?

UPDATE
Thanks for quick responses, +1 to everyone for solution with SCOPE_IDENTITY.
That’s probably my fault, I should have asked the question clearly – I do use MERGE (an example would be much longer, so I posted INSERT instead) , not INSERT so SCOPE_IDENTITY doesn’t really work for me.

Answer 1

A bit shorter version than nesting in a insert statement is using output...into.

declare @tmp_table table(actiontaken nvarchar(10), id int);

merge my_table
using (values ('test')) as S(data)
on 0=1
when not matched then 
  insert (data) values (S.data)
output $action, inserted.id into @tmp_table;

I do believe that you should use a table variable from the merge. The output may contain more than one row.