Question

I have googled this a little and didn’t really find the answer I needed.

I am working on a webpage in C# with SQL Server and LINQ for a customer.
I want the users to be able to send messages to each other. So what I do is that I unit test this with data that actually goes into the database.

The problem is that I now depend on having at least 2 users who I know the ID of. Furthermore I have to clean up after my self. This leads to rather large unit tests that test alot in one test.

Lets say I would like to update a user. That would mean that I would have to ceate the user, update it, and then delete it. This a lot of assertions in one unit test and if it fails with updating i have to manually delete it.

If I would do it any other way, without saving the data to DB, I would not for sure be able to know that the data was present in the database after updating etc.

What is the proper way to do this without having a test that tests a lot of functionality in one test?

Answer 1

Do all the test inside of a System.Transactions.TransactionScope block, and simply do not call Scope.Complete() … All changes will be rolled back when you exit, and other users will not be able to see the temporary data you create in the database, so the test process will only affect the test machine..

You can enter a new user, read the database to verify that it was entered correctly, and whatever else you need to check, within the block of the TransactionScope…

Have the single unit test method impersonate the two different users, all within the one transaction …

e.g. code

   using (var scop = new System.Transactions.TransactionScope())
   {
       // all your test code and Asserts that access the database, 
       // writes and reads, from any class, ...
       // to commit at the very end of this block,
       // you would call
       // scop.Complete();  // ..... but don't and all will be rolled back
   }