How to refresh the value of a #variable from stored procedure resultset between consecutive statements? - sql

Question

Trying to perform multiple consecutive inserts in a table without identity key.
The unique id comes from a procedure called GetNextObjectId. GetNextObjectId is a stored procedure that has no output parameter and no return value.
Instead it selects a top 1 int field.
Tried this:
declare #nextid int;
exec #nextid = GetNextObjectId 1; insert into MyTable values (#nextid, ...);
exec #nextid = GetNextObjectId 1; insert into MyTable values (#nextid, ...);
go
Then this:
declare #nextid int; exec #nextid = GetNextObjectId 1; insert into MyTable values (#nextid, ...);
go
declare #nextid int; exec #nextid = GetNextObjectId 1; insert into MyTable values (#nextid, ...);
go
But the value of #nextid in the insert is always the same.
Question
What is the proper way to refresh the value of this variable without modifying the stored procedure?
Some context
The origin of this question is me looking for a quick way to insert test data in a table using the existing stored procedure, and not managing to do it. The question only relates to the fact the value of the variable does not get updated between statements, not to the proper way to insert data in a table. This is not production code. Also as I understand it, such a procedure is required using Entity Framework with concurrent code; as there are issues with Identity, each thread gets its own ids before saving the context as follows:
// Receive a batch of objects and persist in database
// using Entity Framework.
foreach (var req in requests)
{
// some validation
ctx.MyTable.Add(new Shared.Entities.MyTableType
{
Id = ctx.GetNextObjectId(Enums.ObjectTypes.MyTableType),
Code = req.Code,
Name = req.Name
});
// save to database every 1000 records
counter++;
if (counter % 1000 == 0)
{
ctx.SaveChanges();
counter = 0;
}
}
// save remaining if any
ctx.SaveChanges();
The procedure does this:
BEGIN TRAN T1
UPDATE [dbo].[ObjectsIds] WITH (ROWLOCK)
SET NextId = NextId + Increment
WHERE ObjectTypeId = #objectTypeId
SELECT NextId
FROM [dbo].[ObjectsIds]
WHERE ObjectTypeId = #objectTypeId
COMMIT TRAN T1

Answer 1

There are so many things wrong with this approach that a comment is not sufficient.
First, stored procedures return an integer which is occasionally used. When used, this should be a status value indicating success or failure. There is no requirement but that is how even Microsoft describes the value in the documentation. It sounds like your stored procedure is just running a query, not even returning a status value.
Second, using a stored procedure for this purpose means that you have race conditions. That means that even if the code seemed to work, it might not work for concurrent inserts.
Third, your code is requiring calling a stored procedure as part of every insert. That seems very dangerous, if you actually care about the value.
Fourth, you should be validating the data integrity using a unique index or constraint to prevent subsequent inserts with the same value.
What is the right solution? Well, the best solution is to simply enumerate every row with an identity() column. If you need to do specific counts by a column, then you can calculate that during querying.
If that doesn't meet your needs (although it has always been good enough for me), you can write a trigger. When writing a trigger, you need to be careful about locking the table to be sure that concurrent inserts don't produce the same value. That could suggest using a mechanism such as multiple sequences. Or it could suggest clustering the table around the groups.
The short message: triggers are the mechanism to do what you want (affect the data during a DML operation). Stored procedures are not the mechanism.