Quick Start for Mocking with Typemock Isolator

net comments edit

Introduction

This quick start gets you up to speed on the basic functionality of Typemock Isolator. As you go through it, note:

  • While some basic unit testing will be reviewed, this is not a quick start for NUnit or unit testing in general, so proper test design and specifics on NUnit usage will not be addressed.
  • The quick start shows usage of the Enterprise features of Typemock Isolator. While all of this is possible in the community edition, the alternate mechanisms for accomplishing things won’t be discussed.

At the very end are some take-away points and additional resources. Work through the quick start, do the exercises, and at the end check out the take-away points so you’ll have something to move forward with.

You can get the source for the finished solution here (minus the answers to the exercises - that’s for the reader), but it’s recommended you create your solution and walk through the work and not just get the finished deal: [View Complete Source]

Required Tools

You’ll need to have the following to work through the quick start:

Visual Studio 2008, Typemock Isolator, and TestDriven.NET get installed on your local developer machine. NUnit should be in an accessible location but doesn’t necessarily have to be installed. Get the latest available versions.

Create Test Solution

You’ll need a place where you can run through these exercises, so…

  1. In Visual Studio, select “File -> New Project…”
  2. In the “New Project” dialog…
    1. In the tree view on the left, under “Visual C#” select “Windows.”
    2. In the “Templates” section on the right, select “Class Library.”
    3. Give your library a name like “TypeMockQuickStart.”
  3. In the Solution Explorer, in your new class library…
    1. Delete “Class1.cs” - we’ll add more appropriately named classes later.
    2. Add references to NUnit and Typemock Isolator.
      1. Right-click the “References” folder and select “Add Reference…”
      2. In the .NET tab, select “Typemock Isolator for .NET 2.0” and click OK.
      3. Right-click the “References” folder and select “Add Reference…”
      4. Reference NUnit. If you’ve installed NUnit, select the NUnit Framework assembly from the .NET tab; if you’re accessing NUnit from a known location, go to the Browse tab and find “nunit.framework.dll” and add a reference to that.

Patterns

This section discusses basic patterns you’ll need to understand while working with unit testing and mocking.

Unit Tests: Setup, Execute, Assert

The basic pattern for a unit test is “Setup, Execute, Assert”:

  • Setup: Set up the test environment and the code that you’re testing. This usually involves initializing some variables, instantiating your class, or setting up some configuration files that the code being tested needs.
  • Execute: Execute the code being tested.
  • Assert: Check to make sure that what you just executed had the desired behavior.

To examine this pattern, set up a simple class that can be tested and perform some unit tests on it.

  1. Add a class called Calculator. Make it public.
  2. In the Calculator class, add a method with this signature that adds two doubles (a + b): public double Add(double a, double b)
  3. Fill in the body for the Calculator.Add method.
  4. In the Calculator class, add a method with this signature that divides two doubles (a / b): public double Divide(double a, double b)
  5. Fill in the body for the Calculator.Divide method. If “b” is zero, throw a DivideByZeroException.

Now you have a simple class to test, let’s add a test fixture for it.

  1. Add a class called CalculatorFixture. Make it public.
  2. Add an NUnit [TestFixture] attribute to CalculatorFixture. This is how you tell NUnit and TestDriven.NET that this class contains unit tests.

You now have a class to test and a fixture to contain your tests. Add a test for the Calculator.Add method.

  1. In your test fixture class, add a public void method that takes no parameters called “AddTwoPositiveNumbers”: public void AddTwoPositiveNumbers()
  2. Add an NUnit [Test] attribute to the AddTwoPositiveNumbers method. This is how you tell NUnit and TestDriven.NET that this is a test to run.
  3. Follow the “Setup, Execute, Assert” pattern to create your test.
    1. Setup: Create an instance of the Calculator class.
    2. Execute: Call the Add method on the instance with two positive numbers of your choosing.
    3. Assert: Verify that it returned the expected result.

Your test fixture will look something like this:

[TestFixture]
public class CalculatorFixture
{
  [Test]
  public void AddTwoPositiveNumbers()
  {
    // Setup
    Calculator calc = new Calculator();

    // Execute
    double result = calc.Add(3, 7);

    // Assert
    Assert.AreEqual(10, result);
  }
}

Exercise - Add Tests

Add more tests.

  • Test the Add method:
    • Test adding of two negative numbers.
    • Test adding a number to zero.
  • Test the Divide method:
    • Test dividing one positive number by another.
    • Test dividing by zero (don’t catch the exception - use the NUnit [ExpectedException] attribute on your test method).

Notice how the pattern is basically the same? You do some initialization, run some code, and assert that the result is what you expected. You see a slight deviation from that pattern when testing for expected exceptions, but it’s still basically doing an assertion, just expressed differently.

Mocking: Record, Playback, Verify

The basic pattern for mocking an object is “Record, Playback, Verify”:

  • Record: Tell the mocking framework what you’re about to do on the mock object.
  • Playback: As you’re executing the test, the mocking framework “repeats” what you recorded.
  • Verify: Check to make sure that all the mocks you set up were called correctly.

This parallels the unit test “Setup, Execute, Assert” pattern. Part of your setup is to record your mocks; part of execution is playing back your mocks; part of assertion is verifying your mocks.

Add a new method to the Calculator class that has additional complexity. We’ll use mocking to test this method:

public double AddThenMultiply(double a, double b, double c)
{
  double addResult = this.Add(a, b);
  double multiplyResult = addResult * c;
  return multiplyResult;
}

Notice in this method that we’re calling the Add method (which we’ve already tested) and performing some additional custom logic that we need to test. This sort of thing is perfect for mocking. We don’t want to re-test the Add method; we want to isolate the logic in the new method and just test that. Let’s add a new fixture with tests that use Typemock Isolator to isolate the new logic.

  1. Add a class called CalculatorMockingFixture. Make it public.
  2. Add an NUnit [TestFixture] attribute to CalculatorMockingFixture.
  3. Add a Typemock Isolator [VerifyMocks] attribute to CalculatorMockingFixture. This tells Typemock Isolator to automatically do the mock verification part of the test for you when the test is complete. It saves you from having to manually verify in every test.

The empty fixture should look like this:

[TestFixture]
[VerifyMocks]
public class CalculatorMockingFixture
{
}

Now we’re ready to add a test. In the CalculatorMockingFixture add a test method called MultiplyPositiveAddResult. Here is the method body:

[Test]
public void MultiplyPositiveAddResult()
{
  Calculator calc = new Calculator();
  using (RecordExpectations recorder = RecorderManager.StartRecording())
  {
    double dummy = calc.Add(0, 0);
    recorder.Return((double)15);
  }
  double result = calc.AddThenMultiply(5, 10, 20);
  Assert.AreEqual(300, result);
}

There’s a lot happening here, so let’s look over it:

  1. The Calculator object we’re going to test gets instantiated.
  2. A using block is created where a new RecordExpectations object is created. This is a very common block you will see in Typemock Isolator usage that says, “Everything in this block is fake! Record it and get ready to play it back.” Inside that block…
    1. We’re calling the Add method on the Calculator. Notice that we’re passing 0 for both parameters. The reason for this is that the Add method really isn’t getting called, so it doesn’t matter what we pass. The important part of this is that we’re telling the recorder, “I’m going to call the Add method.” We don’t even really care what we name the variable where we store the result because we’ll never use it - hence, we’ll just call it “dummy.”
    2. We tell the recorder the value we want Add to return - in this case, we want it to return 15. We don’t care what gets passed in, the first time Add gets called, we want 15 to come back.
  3. The using block closes, meaning we’re done recording for now. Time for playback.
  4. We call the AddThenMultiply method and get the result.
  5. We assert that we got the correct result.

What did mocking get us? Try this: change this line:

double result = calc.AddThenMultiply(5, 10, 20);

To this:

double result = calc.AddThenMultiply(500, 1000, 20);

Now run the test again. Notice how it still passes? Why is that? That’s mocking in action. If you follow the call stack, you know that the first thing the AddThenMultiply method does is call the Add method. Typemock Isolator sees that call to Add and doesn’t actually let Add execute - instead, it returns the value we told it to return. In this case, we’ll get 15 back.

Exercise - Experiment

Do some experimentation…

  • What happens if you call Add a second time in your test? Try adding a call to calc.Add after the AddThenMultiply call.
  • Why does that happen?
  • Can you set it up so the recorder returns 15 for the first call to Add but 25 for the second call?

Practical Application

Once you’ve gotten your first mock down and get the patterns, the next question is, “How can I actually use this in my job? I’m not writing Calculator classes all day.” The benefit of mocking is to isolate the code that you’re testing - that includes isolating it from the behavior of the .NET framework and other third-party dependencies. In this section you’ll walk through an example of isolating your code from .NET proper and look at some additional mocking verification that can be done to ensure your code is calling the framework correctly.

Add a Class That Uses Configuration

To experiment with isolation from .NET, we’ll add a new class. This class will make use of the .NET configuration system to read a value from configuration and perform an action based on that.

  1. Add a reference to the System.Configuration assembly.
  2. Add a class called ConfigReader. Make it public.
  3. Add a public method called AppendValueToSetting that takes in a string and returns a string: public string AppendValueToSetting(string valueToAppend)
  4. Fill in the AppendValueToSetting method:
    1. Read the AppSettings key “configReader” and store the result.
    2. Append the contents of the valueToAppend parameter to the end of the value from configuration.
    3. Return the concatenation results.

It should look like this:

public class ConfigReader
{
  public string AppendValueToSetting(string valueToAppend)
  {
    string setting = ConfigurationManager.AppSettings["configReader"];
    string result = setting + valueToAppend;
    return result;
  }
}

Test Cases

Consider what you need to test about this method:

  • What happens if the setting isn’t found?
  • What happens if the setting is empty?
  • What happens if the setting is found?

Now consider: It reads from the App.config file - how do you change that between tests? Is that even a good idea?

Again - mocking to the rescue.

Isolate Yourself From the Framework

  1. Add a test fixture for testing this class. Call the fixture ConfigReaderFixture and include both the [TestFixture] attribute and [VerifyMocks] attribute.
  2. Add a unit test to the fixture. Call it SettingFound - we’ll test what happens when the setting is correctly read from configuration.
  3. In the test…
    1. Create an instance of ConfigReader.
    2. Create a recorder block. Inside the recorder block…
      1. Read the ConfigurationManager.AppSettings["configReader"] key.
      2. Tell the recorder to return the value “readFromAppSettings”.
    3. Call the AppendValueToSetting method and pass in “PassedInFromTest”.
    4. Assert that the value you get back is “readFromAppSettingsPassedInFromTest” - the result of concatenating the two strings.

The test will look like this:

[Test]
public void SettingFound()
{
  ConfigReader reader = new ConfigReader();
  using (RecordExpectations recorder = RecorderManager.StartRecording())
  {
    string dummySetting = ConfigurationManager.AppSettings["configReader"];
    recorder.Return("readFromAppSettings");
  }
  string result = reader.AppendValueToSetting("PassedInFromTest");
  Assert.AreEqual("readFromAppSettingsPassedInFromTest", result);
}

Notice how you didn’t actually have to put anything in App.config - you don’t really want to re-test the functionality of the .NET framework, you just want to test that your code is correct.

Exercise - More Practical Ideas

Take this to the next step…

  • Add a test where the app settings value is null to simulate what happens when it’s not found.
  • Add a test where the app settings value is empty string to simulate when the key is there but has an empty configured value.

Verifying Mock Behavior

Try this: Go to your ConfigReader class and modify this line:

string setting = ConfigurationManager.AppSettings["configReader"];

To be this:

string setting = ConfigurationManager.AppSettings["theWrongKey"];

Now run your tests - they all still pass. Why?

As mentioned earlier, Typemock Isolator will, by default, just notice which methods and properties you’re using, not what the values of parameters are. Sometimes you’ll want to make sure that not only does the third-party dependency return an expected/recorded result but also that your code is passing in the proper parameter values. In this case, you want to not only make sure that you’re getting an expected value from the configuration system but also that your code is asking for the value you think it’s asking for. You’ll want the recorder to “check arguments.”

Leave the wrong config key in your ConfigReader class - we’ll catch that it’s wrong inside the unit tests.

In the SettingFound test, in your recorder block immediately after the call to ConfigurationManager.AppSettings, tell the recorder to check arguments:

[Test]
public void SettingFound()
{
  ConfigReader reader = new ConfigReader();
  using (RecordExpectations recorder = RecorderManager.StartRecording())
  {
    string dummySetting = ConfigurationManager.AppSettings["configReader"];
    recorder.CheckArguments();
    recorder.Return("readFromAppSettings");
  }
  string result = reader.AppendValueToSetting("PassedInFromTest");
  Assert.AreEqual("readFromAppSettingsPassedInFromTest", result);
}

That tells the recorder that you want to ensure the arguments in the mocked statement are correct. In this case, the argument is the string “configReader” that gets passed to ConfigurationManager.AppSettings. Now run the tests: The SettingFound test fails with an exception like this:

TestCase 'TypeMockQuickStart.ConfigReaderFixture.SettingFound'
failed: Typemock Isolator.VerifyException :
Typemock Isolator Verification: Call to System.Collections.Specialized.NameValueCollection.get_Item() Parameter: 1
 expected: <"configReader">
  but was: <"theWrongKey">

Typemock Isolator caught that the wrong parameter was passed - the one in the test (“configReader”) was expected, but your production code used the wrong value (“theWrongKey”). Fix the ConfigReader class back to the correct settings key and your test will pass again.

Mocking Instances

Once you get past the simplest of cases, you start needing to mock specific instances of classes and sometimes you need to mock methods on instances that get created inside non-test code. Typemock Isolator can do both of these things.

Mocking a Current Instance

This exercise will show you how to create an instance of an object where the constructor of the object is mocked and you control the entire life of that object.

Add a new class to your project. This class will use the Calculator class you added earlier. First, add a couple of constructor overloads to your Calculator class:

public class Calculator
{
  private bool _allowAdd;
  public Calculator() : this(true) { }
  public Calculator(bool allowAdd)
  {
    this._allowAdd = allowAdd;
  }
  // Add, Divide, and AddThenMultiply methods omitted for clarity.
}

Now update the Calculator.Add method so if the _allowAdd member variable is false, the Add method won’t run:

public double Add(double a, double b)
{
  if (!this._allowAdd)
  {
    throw new InvalidOperationException("Add operation is not allowed.");
  }
  return a + b;
}

Running your tests, they should all still pass. Notice the default value for allowing the Add operation is true and it’s set in the default constructor. Now in your CalculatorMockingFixture test fixture, add a new test:

[Test]
public void SkipConstructor()
{
  Calculator calc;
  using (RecordExpectations recorder = RecorderManager.StartRecording())
  {
    calc = new Calculator();
  }
  calc = new Calculator();
  double result = calc.Add(2, 2);
  Assert.AreEqual(4, result);
}

This test runs the constructor for the object inside the recorder block - that means the constructor itself will be mocked. In this case, think about what that means - the default value for an uninitialized Boolean is false, so the _allowAdd value, which normally gets initialized to true in the default constructor, will remain false and won’t let the Add operation run.

Run the test and the test will fail:

TestCase 'TypeMockQuickStart.CalculatorMockingFixture.SkipConstructor'
failed: System.InvalidOperationException : Add operation is not allowed.

You can fix the test so it passes by adding an [ExpectedException] attribute to the test and expecting an InvalidOperationException.

Usually constructor logic is more complex than this - it might read from configuration, try to initialize a file in the filesystem, or do some other actions that you may want to control. In cases like these, you may need to mock the constructor for the object.

Exercise - Mock a Current Instance

Add a test where you mock the constructor logic but also test the AddThenMultiply method. What does that look like? How does it differ from the original test you ran with AddThenMultiply?

Mocking a Future Instance

This exercise will show you how to set up expectations on an object that gets created in code you don’t control.

Add a new class called BackwardsCalculator to your project. The constructor of the BackwardsCalculator class will create an instance of Calculator and hang onto it. A method ReverseAdd will perform an add operation using the Calculator and reverse the string value of that. The BackwardsCalculator class looks like this:

public class BackwardsCalculator
{
  private Calculator _calculator;
  public BackwardsCalculator()
  {
    this._calculator = new Calculator();
  }
  public string ReverseAdd(double a, double b)
  {
    string forward = this._calculator.Add(a, b).ToString();
    string reversed = new string(forward.ToCharArray().Reverse().ToArray());
    return reversed;
  }
}

If you feed in 10 and 25 to the ReverseAdd method, you’ll get “53” returned: 10 + 25 is 35, reverse 35 is 53.

Notice how the BackwardsCalculator constructor is creating a Calculator but you don’t get a chance to insert a mock yourself. That’s okay - you can still use mocks to mock out the call to Calculator.Add so you’re isolating your code.

  1. Add a test fixture for testing this class. Call the fixture BackwardsCalculatorFixture and include both the [TestFixture] attribute and [VerifyMocks] attribute.
  2. Add a unit test to the fixture. Call it ReverseAddPositive - we’ll test two positive inputs.
  3. In the test…
    1. Create a recorder block. Inside the recorder block…
      1. Create an instance of the Calculator class.
      2. Call the Add operation and pass in two known values.
      3. Check the arguments on the Add operation to make sure you’re getting called with the expected parameters.
      4. Return the known good result.
    2. Create an instance of BackwardsCalculator.
    3. Call the ReverseAdd method and pass in the parameters you decided on in the record block.
    4. Assert that the value you get back the reversed sum of the two numbers.

The test will look something like this:

[Test]
public void ReverseAddPositive()
{
  using (RecordExpectations recorder = RecorderManager.StartRecording())
  {
    Calculator dummyCalc = new Calculator();
    dummyCalc.Add(10, 20);
    recorder.CheckArguments();
    recorder.Return((double)30);
  }
  BackwardsCalculator bCalc = new BackwardsCalculator();
  string result = bCalc.ReverseAdd(10, 20);
  Assert.AreEqual("03", result);
}

By adding the constructor call to the mock recorder block, we’re saying, “Mock the next instance of this object that gets created and set these expectations on that object.” When we instantiate the BackwardsCalculator after the recorder block, the mocking framework goes into playback mode and mocks the creation of the Calculator in the BackwardsCalculator constructor.

Exercise - Mock a Future Instance

  • If Add was doing more than just adding, this would be perfect for isolating our code. Add is a simple method, though. Look at the CallOriginal method on the recorder and see if you can get a test to pass where you’re checking arguments but actually calling a live version of the Calculator.Add method.
  • The ReverseAdd method is using LINQ to reverse the string. Try mocking the LINQ Reverse or ToArray statements.
  • Look at the WhenArgumentsMatch method on the recorder. It lets you conditionally mock a statement based on the arguments that get passed into a method. Can you set up the mock so it only runs when the arguments match what is getting passed in?

Advanced Mocking

Mocking lets you isolate yourself from a lot of common situations, including:

  • Class factories - Instantiate a class based on a configured value. Isolate yourself from configuration when testing the factory or isolate yourself from the factory when testing code that uses the factory.
  • Third-party dependencies - Components from external vendors. Isolate yourself from the behavior of the dependency so you’re not testing the component, you’re testing your code.
  • .NET framework - Built-in framework classes. Isolate yourself from the internal workings of the framework and test just how you’re using it.
  • Legacy API - Code you have to interact with on legacy systems. Isolate your interaction code from the legacy system so you’re not testing the system.

There are other cases where you might need to use Typemock Isolator, too, like adding tests for an API that was already written and can’t change. You can mock lots of things that can make testing tricky…

  • Fields.
  • Static methods/properties.
  • Private methods/properties.
  • Sealed classes.

Take-Away

If you forget everything else, remember these rules of thumb:

  • Record, Playback, Verify.
  • If it’s in a mock recorder block, it’s not actually running - it’s recording to be played back later, like a tape recorder.
  • Only mock what you need. You can get into some crazy situations by mocking too much, and you might end up in a spot where you’re just testing your mocks and no actual code.
  • Just because you can mock it doesn’t mean you should. For example, Typemock Isolator can mock fields, but you should really stop and look at what you’re mocking before you get too deep. Knowing too much about the internal implementation of a class will make your tests very brittle if anything has to change.
  • You still need to design your application with good principles. Typemock Isolator enables you to test poorly designed code, but that’s not a green light to stop proper software design.

Additional Resources

The Typemock Learn page has a great set of resources, including:

  • API documentation
  • Examples
  • Multimedia
  • Cheat sheets
  • FAQ

The Cheat Sheets are particularly helpful - one-page references you can print off and use while you’re working.

If you use Snippet Compiler, you can experiment with Typemock Isolator using a special template for Snippet Compiler.

Comments