
- •Practical Unit Testing with JUnit and Mockito
- •Table of Contents
- •About the Author
- •Acknowledgments
- •Preface
- •Preface - JUnit
- •Part I. Developers' Tests
- •Chapter 1. On Tests and Tools
- •1.1. An Object-Oriented System
- •1.2. Types of Developers' Tests
- •1.2.1. Unit Tests
- •1.2.2. Integration Tests
- •1.2.3. End-to-End Tests
- •1.2.4. Examples
- •1.2.5. Conclusions
- •1.3. Verification and Design
- •1.5. Tools Introduction
- •Chapter 2. Unit Tests
- •2.1. What is a Unit Test?
- •2.2. Interactions in Unit Tests
- •2.2.1. State vs. Interaction Testing
- •2.2.2. Why Worry about Indirect Interactions?
- •Part II. Writing Unit Tests
- •3.2. Class To Test
- •3.3. Your First JUnit Test
- •3.3.1. Test Results
- •3.4. JUnit Assertions
- •3.5. Failing Test
- •3.6. Parameterized Tests
- •3.6.1. The Problem
- •3.6.2. The Solution
- •3.6.3. Conclusions
- •3.7. Checking Expected Exceptions
- •3.8. Test Fixture Setting
- •3.8.1. Test Fixture Examples
- •3.8.2. Test Fixture in Every Test Method
- •3.8.3. JUnit Execution Model
- •3.8.4. Annotations for Test Fixture Creation
- •3.9. Phases of a Unit Test
- •3.10. Conclusions
- •3.11. Exercises
- •3.11.1. JUnit Run
- •3.11.2. String Reverse
- •3.11.3. HashMap
- •3.11.4. Fahrenheits to Celcius with Parameterized Tests
- •3.11.5. Master Your IDE
- •Templates
- •Quick Navigation
- •Chapter 4. Test Driven Development
- •4.1. When to Write Tests?
- •4.1.1. Test Last (AKA Code First) Development
- •4.1.2. Test First Development
- •4.1.3. Always after a Bug is Found
- •4.2. TDD Rhythm
- •4.2.1. RED - Write a Test that Fails
- •How To Choose the Next Test To Write
- •Readable Assertion Message
- •4.2.2. GREEN - Write the Simplest Thing that Works
- •4.2.3. REFACTOR - Improve the Code
- •Refactoring the Tests
- •Adding Javadocs
- •4.2.4. Here We Go Again
- •4.3. Benefits
- •4.4. TDD is Not Only about Unit Tests
- •4.5. Test First Example
- •4.5.1. The Problem
- •4.5.2. RED - Write a Failing Test
- •4.5.3. GREEN - Fix the Code
- •4.5.4. REFACTOR - Even If Only a Little Bit
- •4.5.5. First Cycle Finished
- •‘The Simplest Thing that Works’ Revisited
- •4.5.6. More Test Cases
- •But is It Comparable?
- •Comparison Tests
- •4.6. Conclusions and Comments
- •4.7. How to Start Coding TDD
- •4.8. When not To Use Test-First?
- •4.9. Should I Follow It Blindly?
- •4.9.1. Write Good Assertion Messages from the Beginning
- •4.9.2. If the Test Passes "By Default"
- •4.10. Exercises
- •4.10.1. Password Validator
- •4.10.2. Regex
- •4.10.3. Booking System
- •Chapter 5. Mocks, Stubs, Test Spies
- •5.1. Introducing Mockito
- •5.1.1. Creating Test Doubles
- •5.1.2. Expectations
- •5.1.3. Verification
- •5.1.4. Conclusions
- •5.2. Types of Test Double
- •5.2.1. Code To Be Tested with Test Doubles
- •5.2.2. The Dummy Object
- •5.2.3. Test Stub
- •5.2.4. Test Spy
- •5.2.5. Mock
- •5.3. Putting it All Together
- •5.4. Example: TDD with Test Doubles
- •5.4.2. The Second Test: Send a Message to Multiple Subscribers
- •Refactoring
- •5.4.3. The Third Test: Send Messages to Subscribers Only
- •5.4.4. The Fourth Test: Subscribe More Than Once
- •Mockito: How Many Times?
- •5.4.5. The Fifth Test: Remove a Subscriber
- •5.4.6. TDD and Test Doubles - Conclusions
- •More Test Code than Production Code
- •The Interface is What Really Matters
- •Interactions Can Be Tested
- •Some Test Doubles are More Useful than Others
- •5.5. Always Use Test Doubles… or Maybe Not?
- •5.5.1. No Test Doubles
- •5.5.2. Using Test Doubles
- •No Winner So Far
- •5.5.3. A More Complicated Example
- •5.5.4. Use Test Doubles or Not? - Conclusion
- •5.6. Conclusions (with a Warning)
- •5.7. Exercises
- •5.7.1. User Service Tested
- •5.7.2. Race Results Enhanced
- •5.7.3. Booking System Revisited
- •5.7.4. Read, Read, Read!
- •Part III. Hints and Discussions
- •Chapter 6. Things You Should Know
- •6.1. What Values To Check?
- •6.1.1. Expected Values
- •6.1.2. Boundary Values
- •6.1.3. Strange Values
- •6.1.4. Should You Always Care?
- •6.1.5. Not Only Input Parameters
- •6.2. How to Fail a Test?
- •6.3. How to Ignore a Test?
- •6.4. More about Expected Exceptions
- •6.4.1. The Expected Exception Message
- •6.4.2. Catch-Exception Library
- •6.4.3. Testing Exceptions And Interactions
- •6.4.4. Conclusions
- •6.5. Stubbing Void Methods
- •6.6. Matchers
- •6.6.1. JUnit Support for Matcher Libraries
- •6.6.2. Comparing Matcher with "Standard" Assertions
- •6.6.3. Custom Matchers
- •6.6.4. Advantages of Matchers
- •6.7. Mockito Matchers
- •6.7.1. Hamcrest Matchers Integration
- •6.7.2. Matchers Warning
- •6.8. Rules
- •6.8.1. Using Rules
- •6.8.2. Writing Custom Rules
- •6.9. Unit Testing Asynchronous Code
- •6.9.1. Waiting for the Asynchronous Task to Finish
- •6.9.2. Making Asynchronous Synchronous
- •6.9.3. Conclusions
- •6.10. Testing Thread Safe
- •6.10.1. ID Generator: Requirements
- •6.10.2. ID Generator: First Implementation
- •6.10.3. ID Generator: Second Implementation
- •6.10.4. Conclusions
- •6.11. Time is not on Your Side
- •6.11.1. Test Every Date (Within Reason)
- •6.11.2. Conclusions
- •6.12. Testing Collections
- •6.12.1. The TDD Approach - Step by Step
- •6.12.2. Using External Assertions
- •Unitils
- •Testing Collections Using Matchers
- •6.12.3. Custom Solution
- •6.12.4. Conclusions
- •6.13. Reading Test Data From Files
- •6.13.1. CSV Files
- •6.13.2. Excel Files
- •6.14. Conclusions
- •6.15. Exercises
- •6.15.1. Design Test Cases: State Testing
- •6.15.2. Design Test Cases: Interactions Testing
- •6.15.3. Test Collections
- •6.15.4. Time Testing
- •6.15.5. Redesign of the TimeProvider class
- •6.15.6. Write a Custom Matcher
- •6.15.7. Preserve System Properties During Tests
- •6.15.8. Enhance the RetryTestRule
- •6.15.9. Make an ID Generator Bulletproof
- •Chapter 7. Points of Controversy
- •7.1. Access Modifiers
- •7.2. Random Values in Tests
- •7.2.1. Random Object Properties
- •7.2.2. Generating Multiple Test Cases
- •7.2.3. Conclusions
- •7.3. Is Set-up the Right Thing for You?
- •7.4. How Many Assertions per Test Method?
- •7.4.1. Code Example
- •7.4.2. Pros and Cons
- •7.4.3. Conclusions
- •7.5. Private Methods Testing
- •7.5.1. Verification vs. Design - Revisited
- •7.5.2. Options We Have
- •7.5.3. Private Methods Testing - Techniques
- •Reflection
- •Access Modifiers
- •7.5.4. Conclusions
- •7.6. New Operator
- •7.6.1. PowerMock to the Rescue
- •7.6.2. Redesign and Inject
- •7.6.3. Refactor and Subclass
- •7.6.4. Partial Mocking
- •7.6.5. Conclusions
- •7.7. Capturing Arguments to Collaborators
- •7.8. Conclusions
- •7.9. Exercises
- •7.9.1. Testing Legacy Code
- •Part IV. Listen and Organize
- •Chapter 8. Getting Feedback
- •8.1. IDE Feedback
- •8.1.1. Eclipse Test Reports
- •8.1.2. IntelliJ IDEA Test Reports
- •8.1.3. Conclusion
- •8.2. JUnit Default Reports
- •8.3. Writing Custom Listeners
- •8.4. Readable Assertion Messages
- •8.4.1. Add a Custom Assertion Message
- •8.4.2. Implement the toString() Method
- •8.4.3. Use the Right Assertion Method
- •8.5. Logging in Tests
- •8.6. Debugging Tests
- •8.7. Notifying The Team
- •8.8. Conclusions
- •8.9. Exercises
- •8.9.1. Study Test Output
- •8.9.2. Enhance the Custom Rule
- •8.9.3. Custom Test Listener
- •8.9.4. Debugging Session
- •Chapter 9. Organization Of Tests
- •9.1. Package for Test Classes
- •9.2. Name Your Tests Consistently
- •9.2.1. Test Class Names
- •Splitting Up Long Test Classes
- •Test Class Per Feature
- •9.2.2. Test Method Names
- •9.2.3. Naming of Test-Double Variables
- •9.3. Comments in Tests
- •9.4. BDD: ‘Given’, ‘When’, ‘Then’
- •9.4.1. Testing BDD-Style
- •9.4.2. Mockito BDD-Style
- •9.5. Reducing Boilerplate Code
- •9.5.1. One-Liner Stubs
- •9.5.2. Mockito Annotations
- •9.6. Creating Complex Objects
- •9.6.1. Mummy Knows Best
- •9.6.2. Test Data Builder
- •9.6.3. Conclusions
- •9.7. Conclusions
- •9.8. Exercises
- •9.8.1. Test Fixture Setting
- •9.8.2. Test Data Builder
- •Part V. Make Them Better
- •Chapter 10. Maintainable Tests
- •10.1. Test Behaviour, not Methods
- •10.2. Complexity Leads to Bugs
- •10.3. Follow the Rules or Suffer
- •10.3.1. Real Life is Object-Oriented
- •10.3.2. The Non-Object-Oriented Approach
- •Do We Need Mocks?
- •10.3.3. The Object-Oriented Approach
- •10.3.4. How To Deal with Procedural Code?
- •10.3.5. Conclusions
- •10.4. Rewriting Tests when the Code Changes
- •10.4.1. Avoid Overspecified Tests
- •10.4.2. Are You Really Coding Test-First?
- •10.4.3. Conclusions
- •10.5. Things Too Simple To Break
- •10.6. Conclusions
- •10.7. Exercises
- •10.7.1. A Car is a Sports Car if …
- •10.7.2. Stack Test
- •Chapter 11. Test Quality
- •11.1. An Overview
- •11.2. Static Analysis Tools
- •11.3. Code Coverage
- •11.3.1. Line and Branch Coverage
- •11.3.2. Code Coverage Reports
- •11.3.3. The Devil is in the Details
- •11.3.4. How Much Code Coverage is Good Enough?
- •11.3.5. Conclusion
- •11.4. Mutation Testing
- •11.4.1. How does it Work?
- •11.4.2. Working with PIT
- •11.4.3. Conclusions
- •11.5. Code Reviews
- •11.5.1. A Three-Minute Test Code Review
- •Size Heuristics
- •But do They Run?
- •Check Code Coverage
- •Conclusions
- •11.5.2. Things to Look For
- •Easy to Understand
- •Documented
- •Are All the Important Scenarios Verified?
- •Run Them
- •Date Testing
- •11.5.3. Conclusions
- •11.6. Refactor Your Tests
- •11.6.1. Use Meaningful Names - Everywhere
- •11.6.2. Make It Understandable at a Glance
- •11.6.3. Make Irrelevant Data Clearly Visible
- •11.6.4. Do not Test Many Things at Once
- •11.6.5. Change Order of Methods
- •11.7. Conclusions
- •11.8. Exercises
- •11.8.1. Clean this Mess
- •Appendix A. Automated Tests
- •A.1. Wasting Your Time by not Writing Tests
- •A.1.1. And what about Human Testers?
- •A.1.2. One More Benefit: A Documentation that is Always Up-To-Date
- •A.2. When and Where Should Tests Run?
- •Appendix B. Running Unit Tests
- •B.1. Running Tests with Eclipse
- •B.1.1. Debugging Tests with Eclipse
- •B.2. Running Tests with IntelliJ IDEA
- •B.2.1. Debugging Tests with IntelliJ IDEA
- •B.3. Running Tests with Gradle
- •B.3.1. Using JUnit Listeners with Gradle
- •B.3.2. Adding JARs to Gradle’s Tests Classpath
- •B.4. Running Tests with Maven
- •B.4.1. Using JUnit Listeners and Reporters with Maven
- •B.4.2. Adding JARs to Maven’s Tests Classpath
- •Appendix C. Test Spy vs. Mock
- •C.1. Different Flow - and Who Asserts?
- •C.2. Stop with the First Error
- •C.3. Stubbing
- •C.4. Forgiveness
- •C.5. Different Threads or Containers
- •C.6. Conclusions
- •Appendix D. Where Should I Go Now?
- •Bibliography
- •Glossary
- •Index
- •Thank You!

Chapter 11. Test Quality
See how fast the tests run. If you find it unacceptably long, or notice some pauses during the execution of the tests, then look further, and answer the following questions:
•Are these unit tests, or rather integration tests? Typically, setting up an application context or database connection takes time.
•Are the setup methods used properly? Although this happens rarely with unit tests, do look to see whether some objects might, for example, be created before the class rather than just before each method (see Section 3.8.4).
•Are there any deliberate Thread.sleep() calls? (See the next section for some discussion).
Also, look at the messages printed during the execution of the tests (usually on the console). If you cannot keep up with them, it means you have a Loudmouth issue (see Section 8.5).
Make sure the tests are repeatable. Usually they are, but in cases of multithreading tests (see Section 6.10), or tests with random values (see Section 7.2), this might not be the case.
Date Testing
Experience tells me that the testing of time-dependent business logic is rarely done correctly (see Section 6.11). Common issues you should look for are the following:
•Look for any Thread.sleep() code constructs which make unit tests run longer than is really required.
•A common mistake is to test only the current date, which means the bug will be discovered exactly on the day the code fails in production. In general you need to make sure all test cases are covered.
11.5.3. Conclusions
Of the many approaches to ensuring the quality of your test code, code reviews are to be recommended the most strongly. They help to find bugs, help your team to converge with respect to how they all go about coding, and serve to disseminate knowledge amongst team members about parts of the software being developed. Moreover, by examining the test code, a lot can be discovered about production code. Frankly, what more could one ask for? The only downside is that to perform a thorough examination of test code, a lot of time is required.
I strongly recommend making test code reviews a part of your team’s working routine. Code reviews should belong to your Definition of Done, and they should embrace test code in the same manner that they embrace production code.
11.6. Refactor Your Tests
God grant me serenity to accept the code I cannot change, courage to change the code I can, and wisdom to know the difference.
— Erik Naggum
Now that we know in what ways we can measure test quality, it is time to actually fix the deficiencies uncovered by code coverage (see Section 11.3), mutation testing (see Section 11.4) and code reviews
254

Chapter 11. Test Quality
(see Section 11.5). In this section we discuss the process of refactoring, which is a common way of altering code. We have already discussed a refactoring in Chapter 4, Test Driven Development, as one of the phases of the TDD rhythm. In this section we shall discuss the refactorings that are common when working with tests. Some of them are identical to the ones we perform on production code, but some are specific to test code.
Numerous examples of making improvements to test code by introducing changes of various sorts have been given already in other parts of this book – or, indeed, are easily deducible from what has been discussed earlier. For example, we already know about the advantages of using assertEquals() over assertTrue() for comparing objects, and we know what the advantages of matchers are (see Section 6.6): both can be the source of multiple refactorings, which should be used to improve the readability of assertions. Similarly, making use of data providers (instead of for loops), or having specialized setUp() methods, can also serve as a refactoring hints. This section supplements what has already been discussed with new examples and approaches that have yet to be examined.
Before we see some code, let us think about the reasons for test code refactorings. The first and most significant one is that the test code is very important and should be cleaned so that it is easy to understand and maintain. Refactoring helps you achieve this. By introducing various, sometimes minor, changes to test code, you can make your tests convey better exactly what you are trying to test. If a wellwritten test fails, it is also easier to find out why.
There is also a question about how to refactor test code. The situation is different from that with production code, because we do not have tests for our tests. However, we can deal with this issue by following these simple pieces of advice:
•You should perform refactorings in small steps, and rerun tests often along the way.
•Additional tools - code coverage and mutation testing - may help to find out if your safety net of tests is not getting loosened up.
Some people say that before starting to refactor tests, you should first change the SUT’s implementation, so the test fails21. After refactoring tests and rerunning them, you should still see them failing - which is a sign that the assertions are still working. Now, when you revert the changes in production code, the tests should pass again. I have to admit, I have never used the last technique during any serious refactoring. Moving in small steps worked well enough for me.
Another thing we should discuss is when to refactor a particular test. An obvious moment for this is the refactoring phase of the TDD rhythm, but this is not the only one. In fact, I would encourage you to do it every time you do not feel comfortable when browsing the test code. It does not matter if you have written that code yourself or not. Remember the Boy Scout Rule ("Leave the campground cleaner than you found it!") and dare to make the small changes which, added together, still make a big difference!
Before we see some examples, let me inform you that all of them involve real code. Some names of methods and classes have been changed to "protect the innocent". :)
The examples are only big enough to demonstrate what is to be done, and how the code changes after refactoring. Each of the refactorings shown addresses some real issues within the test code. Some of them might not seem very dangerous when looking at ten lines of test code examples, but their importance
21Please note that this approach is exactly the opposite of what you do when refactoring production code, which should be performed only if all tests are green.
255

Chapter 11. Test Quality
will grow along with the size of your test codebase. Likewise, the benefits of refactorings are much more clearly visible when they are introduced for real-sized test classes. Please bear this in mind when contemplating the examples in this section.
11.6.1. Use Meaningful Names - Everywhere
We have already discussed the importance of good naming. We have discussed various patterns for test class names and test method names, and pondered over the naming schema for test doubles (see Section 9.2). Throughout the book I have been encouraging you to rename things if you do not feel comfortable with the existing names. Good, meaningful names are invaluable!
This section presents another facet of the same issue. It shows how the simple refactoring of renaming variables can make it easier to understand the logic of test code.
Imagine a test of a class from a security module, which is supposed to allow or deny access to users based on their permissions. First the test registered users (identified by string id) and assigned them some permissions. Then it verified that users had only those permissions that had been assigned to them. Listing 11.6 shows a small part of this test.
Listing 11.6. User_1, user_2, user_3 - who are you?
public static Object[] userHasPermissions() { return $(
$("user_1", Permission.READ),
$("user_1", Permission.WRITE),
$("user_1", Permission.REMOVE),
$("user_2", Permission.WRITE),
$("user_2", Permission.READ),
$("user_3", Permission.READ)
);
}
@Parameters(method = "userHasPermissions") @Test
public void shouldReturnTrueIfUserHasPermission(String username, Permission permission) {
assertTrue(sut.hasPermission(username, permission));
}
The problem here is that this test is not obvious. For someone who looks at the test for the first time, it is not obvious whether user_2 should or should not have been granted the READ permission. Who the heck is user_2? Well, this must be checked by analyzing the data of previously registered users (probably in a setUp() method somewhere).
A simple change can achieve wonders. Look at the updated version shown in Listing 11.7.
256

Chapter 11. Test Quality
Listing 11.7. Admin, logged on user, guest - I know who you are!
public static Object[] usersPermissions() { return $(
$(ADMIN, Permission.READ),
$(ADMIN, Permission.WRITE),
$(ADMIN, Permission.REMOVE),
$(LOGGED, Permission.WRITE),
$(LOGGED, Permission.READ), $(GUEST, Permission.READ)
);
}
@Parameters(method = "usersPermissions") @Test
public void shouldReturnTrueIfUserHasPermission(String username, Permission permission) {
assertTrue(sut.hasPermission(username, permission));
}
Now this is clear! Admin should have all possible permissions. A "normal" logged on user should be able to read and write, but not to remove things. A guest user can normally only see what others have uploaded, but cannot change the content himself or herself. There is no need to consult any documentation: the code speaks for itself.
11.6.2. Make It Understandable at a Glance
The readability of tests can be improved in lots of ways. Take a look at the following snippet of test code, which creates an object of the MockServer class:
Listing 11.8. Not easy to understand what type of server is created
server = new MockServer(responseMap, true,
new URL(SERVER_ROOT).getPort(), false);
What properties does a server variable have? What kind of server is created? If you do not remember the API of MockServer, then you need to ask your IDE for help, so it explains the meaning of true and false flags being passed to the MockServer constructor. Would it be possible to change this code so it is easier to understand? Yes, by introducing some values whose names tell a story:
Listing 11.9. Self-explanatory values passed to the MockServer constructor
private static final boolean NO_SSL = false;
private static final boolean RESPONSE_IS_A_FILE = true;
server = new MockServer(responseMap, RESPONSE_IS_A_FILE, new URL(SERVER_ROOT).getPort(), NO_SSL);
Now this makes more sense - this server responds with a file, and does not use SSL.
Another way in which this code could be made more readable is by using the Test Data Builder pattern (see Section 9.6.2).
257