Question

ALTER PROCEDURE [dbo].[spWeb_TransactionProjectUpdate]
    @transactionId              INT,
    @projectId                  INT
AS

IF NOT EXISTS(SELECT * FROM dbo.com_project_transaction_link WHERE pt_tr_transaction_id = @transactionId AND pt_pj_project_id = @projectId)
BEGIN

    INSERT INTO dbo.com_project_transaction_link(pt_tr_transaction_id, pt_pj_project_id) VALUES (@transactionId, @projectId)

END

So when we run this with after an identical record is already in the database, it fails. But we have the IF NOT EXISTS there for this exact reason. Why would it be trying to insert if it finds a record?

Answer 1

If that pair of columns (pt_tr_transaction_id, pt_pj_project_id) is unique, why not just put a UNIQUE INDEX on these two columns? Combine this with the IGNORE_DUP_KEY option on that index, and you would have an index that

• prevents any duplicates from being inserted
• doesn’t barf (throw exception) when a duplicate is encountered – it’s just being swallowed silently

This check of yours is prone to failure, since you’re not doing anything to prevent two clients to running at the same time; both could first check to see if that row exists and get back a false, and then both could proceed to insert those values. The check you have is not safe against concurrent execution.