Month: April 2011

One of the most common gripes that I hear about getting started with unit testing is that the code isn’t testable or it doesn’t have a testable design. People will make claims like, “We need X hours to make the project testable before we can start writing tests.” Well, I don’t buy into that. There are techniques that you can employ to extract the pieces of code you’re modifying into little, testable nuggets that have nothing to do with the rest of the not-so-test-friendly project. I’m going to cover two of my favorite, most-commonly used techniques in this post.

Extract the code to be tested.

This is a pretty straightforward solution to a not-so-obvious problem. The most common scenario that I’ve seen for this technique is the “monster function.” I’m sure you’ve seen this before. It’s the 1000-line function that contains all the logic for the module or application you’re modifying. This is a classic hard-to-test situation because, at first glance, it seems like you need to set up everything necessary to complete the entire process in order to do any sort of unit testing.

To transform this mess into an unit testing masterpiece, you just need to do what nobody before you thought of: create a new function. Your new function can do whatever you need it to do. It can accept inputs. It can return outputs. And, it can do it all in a testable manner! So, make your function, test the hell out of it, then call your function from the appropriate place in the main function. Congratulations–you just made a unit tested change to a previously non-testable function.

Extract the code to be avoided.

The second scenario is a little more difficult, and the solution is a little less obvious. I use this technique when I’m dealing with external dependencies that aren’t enough to warrant an entire interface or when I’m working with tricky internal dependencies. Let’s look at an example. Consider the following function:

public class HardToTest
{
    public void ProcessFile()
    {
        string contents;
        using (var fs = new FileStream(@"c:\data.xml", FileMode.Open))
        {
            var sr = new StreamReader(fs);
            contents = sr.ReadToEnd();
        }
        // Do something interesting with file contents
    }
}

When sitting down to write a unit test for this function, you might be tempted to create a “data.xml” file in the root directory of the C drive during TestInitialize. That would probably result in a passing unit test on your machine, but please, don’t do that. You can trust that the FileStream and StreamReader objects will do their jobs correctly, so let’s just take them out of the equation for our unit testing. With a quick refactor, we end up with something like this:

public class HardToTest
{
    public void ProcessFile()
    {
        string contents = GetFileContents();
        // Do something interesting with file contents
    }
    
    internal protected virtual string GetFileContents()
    {
        using (var fs = new FileStream(@"c:\data.xml", FileMode.Open))
        {
            var sr = new StreamReader(fs);
            return sr.ReadToEnd();
        }
    }
}

Ok, great! Now, here’s where it gets fun. We can write a test for this function by mocking just the function we’re trying to avoid. Check it out:

[TestMethod()]
public void ProcessFileTest()
{
    HardToTest target = MockRepository.GeneratePartialMock();
    target.Expect(x => x.GetFileContents()).Return("File contents!");
    target.ProcessFile();
    target.VerifyAllExpectations();
}

Because we made our GetFileContents function virtual, the generated mock class can override its functionality. The Partial Mocks gives us this functionality; if the function were not overridden, it would be called as-is and would try to read from the hard-coded file path.

You could stick with this same approach to handle different scenarios that might arise in the isolated code. For instance, what would happen if the file did not exist? You would expect the isolated code to throw a System.IO.FileNotFoundException, right? Well, let’s test that!

[TestMethod]
[ExpectedException(typeof(System.IO.FileNotFoundException))]
public void ProcessFile_ThrowsFileNotFoundException()
{
    HardToTest target = MockRepository.GeneratePartialMock();
    target.Expect(x => x.GetFileContents()).Throw(
        new System.IO.FileNotFoundException());
    target.ProcessFile();
}

Some additional notes about this solution:

• The function must be internal (see next note) protected (prevents calling of the function while setting expectation) virtual (allows function to be overridden).
• The target project must expose its internals to the test project. This can be done by adding the “[assembly: InternalsVisibleTo(“YourTestProjectName”)]” attribute to the target project’s assembly.info file. You can read more about this attribute here.

When you’re writing unit tests for your code, you want to write specific tests for specific units. Unfortunately, it is not uncommon for you code to have many external dependencies that may not be related to the functionality that you’re trying to test. This is where one of my very favorite unit testing tools, Rhino Mocks, comes into play.

Rhino Mocks is great because it allows you to mock or stub objects in your code so that you can really focus on the functionality that you’re trying to test without having to worry about doing an excessive amount of setup or external dependencies.

When you’re getting started with Rhino Mocks, one of the most confusing things that you’ll run into is that there are several different syntaxes that can be used to accomplish the same thing. The AAA (Arrange, Act, Assert) syntax is the newest and was introduced with version 3.5. The primary distinction of the AAA syntax is that you don’t need to create and manage a MockRepository object and its various states.

In this article, I will walk you through the task of writing a unit test using Rhino Mocks.

So, before we begin writing our tests, let’s look at the code we’ll be testing.

public class Controller
{
    public IDataSource DataSource { get; set; }

    public int[] Data { get; set; }

    public void Process()
    {
        if (this.DataSource.DataAvailable())
        {
            this.Data = this.DataSource.GetData();
        }
    }
}

We can see that we have a Controller class that utilizes some sort of IDataSource object to check for and retrieve data. When we’re testing our controller class, we may or may not have access to our data source, though. And, even if we do have access to the data source, we may not have complete control over it. So this is an excellent situation to employ Rhino Mocks!

The first step to using Rhino Mocks is simply adding a reference to your project. Once the reference is in place, we can add the using statement to the top of our test class and get to mockin’.

Creating a mock object is extremely simple with Rhino Mocks’s AAA syntax. You just use the static MockRepository class’s GenerateMock function.

var mockDataSource = MockRepository.GenerateMock();

Now that we’ve created our mock object, we need to configure our target object to use it. In this case, our class’s IDataSource is a public property on the class. So configuring the target object is a simple assignment.

Controller target = new Controller();
target.DataSource = mockDataSource;

Okay, great! Our next step is to set the expectations for our mock object. For our first test, we’ll just have the call to DataAvailable return false.

mockDataSource.Expect(x => x.DataAvailable()).Return(false);

The last steps for this test are to run the function we’re testing and verify that our mock object was used as expected.

target.Process();
mockDataSource.VerifyAllExpectations();

When we run this test, the Process function will call our mock object’s DataAvailable function–which we have pre-configured to return false–and the test will pass as long as DataAvailable is called.

We can run this test, and it should pass. Awesome, now let’s write another test. This time, we’ll have DataAvailable return true. We create the mock object the same way as previously, but we’ll need to set the expectations differently.

mockDataSource.Expect(x => x.DataAvailable()).Return(true);
var data = new int[] {1, 2, 3};
mockDataSource.Expect(x => x.GetData()).Return(data);

Since we’re passing some data in, we can also add an extra assertion in our unit test.

mockDataSource.VerifyAllExpectations();
Assert.AreEqual(data, target.Data);

This unit test will fail unless both DataAvailable and GetData are called. Additionally, the target object’s Data property must be the object returned by GetData.

Here is the complete text for the two unit tests we’ve just created.

[TestMethod()]
public void ProcessTest_DataAvailableReturnsFalse()
{
    var mockDataSource = MockRepository.GenerateMock();

    Controller target = new Controller();
    target.DataSource = mockDataSource;

    mockDataSource.Expect(x => x.DataAvailable()).Return(false);

    target.Process();
    mockDataSource.VerifyAllExpectations();
}

[TestMethod()]
public void ProcessTest_DataAvailableReturnsTrue()
{
    var mockDataSource = MockRepository.GenerateMock();

    Controller target = new Controller();
    target.DataSource = mockDataSource;

    mockDataSource.Expect(x => x.DataAvailable()).Return(true);
    var data = new int[] { 1, 2, 3 };
    mockDataSource.Expect(x => x.GetData()).Return(data);

    target.Process();
    mockDataSource.VerifyAllExpectations();
    Assert.AreEqual(data, target.Data);
}