Weird Behavior of Trigger - sql-server

Question

I've got a trigger (SQL 2008 R2) that does one simple operation but the results are not logical.
Here is the overview:
A text file is fed to an SSIS package with ONE line(one record) that loads it into "ORDERS_IN_PROCESS" table. The Data Access Mode is set to "Table or view" in the "OLE DB Destination" to allow triggers to fire.
Here is my ORDERS table:
OrderID ItemNo
--------- ---------
9813 1
9813 2
9813 3
9817 1
So, SSIS executes and
ORDERS_IN_PROCESS gets one record inserted which is OrderID 9813
Trigger is fired:
INSERT INTO ORDERS_ARCHIVE SELECT * FROM ORDERS WHERE OrderID=INSERTED.OrderID
Pretty simple so far...
The results I get in my ORDERS_ARCHIVE (identical layout to ORDERS) Table are
OrderId ItemNo
--------- ----------
9813 3
Where are the rest of the 2 line items?
Note, it only inserted the last row read from ORDERS table into ORDERS_ARCHIVE.
I need all 3 of them in ORDERS_ARCHIVE.
Why does this happen?
I believe it has something to do with the way SSIS processes it using "OLE DB Destination" because if I insert a record into RLFL manually, the trigger does exactly what it's supposed to do and inserts all 3 records from BACK.
You may argue that trigger fires once per batch and I agree but in this case I have a batch of just ONE record.
I'm thinking of an sp, but i'd rather not add another level of complexity for something so trivial, supposedly.
Any ideas?
Thanks.

Answer 1

I don't think that this has something to do with SSIS at all, but with your trigger. Instead of than IN that you have there, try using a JOIN in your query:
INSERT INTO ORDERS_ARCHIVE
SELECT O.*
FROM ORDERS O
INNER JOIN INSERTED I
ON O.ORderID = I.OrderID

Answer 2

I concur with Lamark, misspelling intentional, with the assessment of your trigger is incorrect.
The logic you provided for your trigger does not compile. The WHERE clause is not valid. I'm assuming, as Lamak did that your intention was to join based on OrderID.
create table dbo.ORDERS_ARCHIVE
(
OrderID int
, ItemNo int
)
GO
create table dbo.ORDERS
(
OrderID int
, ItemNo int
)
GO
create trigger
trUpdate
ON
dbo.ORDERS
AFTER INSERT
AS
BEGIN
SET NOCOUNT ON;
-- This doesn't work
-- Msg 4104, Level 16, State 1, Procedure trUpdate, Line 12
-- The multi-part identifier "INSERTED.OrderID" could not be bound.
--INSERT INTO dbo.ORDERS_ARCHIVE
--SELECT *
--FROM ORDERS
--WHERE OrderID=INSERTED.OrderID;
-- I think you meant
INSERT INTO dbo.ORDERS_ARCHIVE
SELECT *
FROM ORDERS
WHERE OrderID=(SELECT INSERTED.OrderID FROM INSERTED);
END
GO
I then ginned up a simple SSIS package, I have a data source that supplies the 4 rows you indicated and writes to dbo.ORDERS. I ran the package 2 times and each one netted 4 rows in the ORDERS_ARCHIVE table. 3 rows with 9813, 1 with 9817 per batch.
I am getting the right count of rows in there so I believe the trigger is firing correctly. Instead, what is happening is the logic is incorrect. Since the OrderID is not unique in the ORDERS table, the database engine is going to pick the first row that happens to satisfy the search criteria. It just so happens that it picks the same row (ItemNo = 1) each time but since there is no guarantee of order without an ORDER BY clause, this is just random or an artifact of how the Engine chooses but no behaviour I would bank on remaining consistent.
How do you fix this?
Fix the trigger. Joining to the inserted virtual table only on the OrderID is resulting in multiple rows satisfying the condition.
create trigger
trUpdate
ON
dbo.ORDERS
AFTER INSERT
AS
BEGIN
SET NOCOUNT ON;
-- This trigger will add all the rows from the ORDERS table
-- that match what was just inserted based on OrderID and ItemNo
INSERT INTO dbo.ORDERS_ARCHIVE
SELECT O.*
FROM dbo.ORDERS O
INNER JOIN INSERTED I
ON O.OrderID = I.OrderID
AND O.ItemNo = I.ItemNo;
END
Now when I run the ETL, I see 4 rows in ORDERS_ARCHIVE with the correct ItemNo values.