What is the best design pattern for preventing duplicate submissions?

Question

I have a design pattern I have been struggling with on how best to prevent duplicate posting of data.

Here are the steps:

1. Client submits data with a unique guid (client generated guid - guaranteed unique)
2. Server side software makes sure client guid doesn't exist yet in the DB
3. begins transaction
4. process data (can take between 1-20 seconds depending upon payload)
5. commits transaction

Here are the scenarios: Client submits data with guid "1", and then resubmits data with guid "1" before step (5) is hit for the original data submission, then the transaction is processed twice.

What is the best design pattern to prevent this without using semaphores or blocking? The user should be able to resubmit, in case the first submission fails for some reason (hardware issue on the server side, etc.).

Thanks!

Answer 1

I don't know what you are using to develop your front end, but in a web application you can use ajax to check on the server transition status giving the user some feedback while waiting and also disabling the submit option.

Answer 2

You could implement step 2 by using a query which reads uncommitted data. For instance, if you are using MS SQL Server, you could do this:

IF NOT EXIST(SELECT * FROM SomeTable (NOLOCK) WHERE Guid = @ClienGUID)
BEGIN
   -- Insert the GUID ASAP in the transaction so that the next query will read it
   -- Do steps 3-5
END

The key here is the (NOLOCK) hint, which reads uncommitted data

Answer 3

Could you hash the data that the user is providing and store it in a table -- check that the hash doesn't match any previous submission before continuing?

Answer 4

Store the GUID in a column with an SQL UNIQUE constraint.

When you attempt (within the transaction) to insert a second duplicate GUID, the operation will fail, at which point you roll back the entire transaction.