Isolating Integration Test Data

Published by Olcay Bayram on 17 December 2017

When we write an integration test, we should leave the persistence as it was before. I will show you how to do it with NUnit easily.

After covering every corners of our code with unit tests, we can move on to integration tests that we normally test happy path scenarious. If we are using a database as a persistence then we need to run our DML commands backwards. So you should write your CRUD operations for DRUC (delete, read, update and create) as well.

On the other hand, we can use good old TransactionScope in the System library. Thanks to NUnit unit-testing framework, we do not need to initialize a transaciton scope every time we write a test case. We can create an attribute which is inherited ITestAction interface so it would be ran before and after every test case.

In the BeforeTest method, it creates a new TransactionScope. In the AfterTest method, instead of calling Complete() function of the transaction scope to commit all the changes to the database, we call Dispose() to rollback the specific transaction.

using NUnit.Framework;
using System;
using System.Transactions;

namespace OtomatikMuhendis.IntegrationTests
{
  public class Isolated : Attribute, ITestAction
  {
    private TransactionScope _transactionScope;

    public ActionTargets Targets
    {
      get { return ActionTargets.Test; }
    }

    public void BeforeTest(TestDetails testDetails)
    {
      _transactionScope = new TransactionScope();
    }

    public void AfterTest(TestDetails testDetails)
    {
      _transactionScope.Dispose();
    }
  }
}

We can apply the attrribute to the test cases that we want to isolate.

[Test, Isolated]
public void Update_WhenCalled_ShouldUpdateTheGivenItem()
{ /* ... */ }