xUnit Test Patterns: Refactoring Test Code (Addison-Wesley Signature Series (Fowler)) (英語) ハードカバー – 2007/5/21
Kindle 端末は必要ありません。無料 Kindle アプリのいずれかをダウンロードすると、スマートフォン、タブレットPCで Kindle 本をお読みいただけます。
Automated testing is a cornerstone of agile development. An effective testing strategy will deliver new functionality more aggressively, accelerate user feedback, and improve quality. However, for many developers, creating effective automated tests is a unique and unfamiliar challenge.
xUnit Test Patterns is the definitive guide to writing automated tests using xUnit, the most popular unit testing framework in use today. Agile coach and test automation expert Gerard Meszaros describes 68 proven patterns for making tests easier to write, understand, and maintain. He then shows you how to make them more robust and repeatable--and far more cost-effective.
Loaded with information, this book feels like three books in one. The first part is a detailed tutorial on test automation that covers everything from test strategy to in-depth test coding. The second part, a catalog of 18 frequently encountered "test smells," provides trouble-shooting guidelines to help you determine the root cause of problems and the most applicable patterns. The third part contains detailed descriptions of each pattern, including refactoring instructions illustrated by extensive code samples in multiple programming languages.
Topics covered include
- Writing better tests--and writing them faster
- The four phases of automated tests: fixture setup, exercising the system under test, result verification, and fixture teardown
- Improving test coverage by isolating software from its environment using Test Stubs and Mock Objects
- Designing software for greater testability
- Using test "smells" (including code smells, behavior smells, and project smells) to spot problems and know when and how to eliminate them
- Refactoring tests for greater simplicity, robustness, and execution speed
This book will benefit developers, managers, and testers working with any agile or conventional development process, whether doing test-driven development or writing the tests last. While the patterns and smells are especially applicable to all members of the xUnit family, they also apply to next-generation behavior-driven development frameworks such as RSpec and JBehave and to other kinds of test automation tools, including recorded test tools and data-driven test tools such as Fit and FitNesse.
Refactoring a Test
PART I: The Narratives
Chapter 1 A Brief Tour
Chapter 2 Test Smells
Chapter 3 Goals of Test Automation
Chapter 4 Philosophy of Test Automation
Chapter 5 Principles of Test Automation
Chapter 6 Test Automation Strategy
Chapter 7 xUnit Basics
Chapter 8 Transient Fixture Management
Chapter 9 Persistent Fixture Management
Chapter 10 Result Verification
Chapter 11 Using Test Doubles
Chapter 12 Organizing Our Tests
Chapter 13 Testing with Databases
Chapter 14 A Roadmap to Effective Test Automation
PART II: The Test Smells
Chapter 15 Code Smells
Chapter 16 Behavior Smells
Chapter 17 Project Smells
PART III: The Patterns
Chapter 18 Test Strategy Patterns
Chapter 19 xUnit Basics Patterns
Chapter 20 Fixture Setup Patterns
Chapter 21 Result Verification Patterns
Chapter 22 Fixture Teardown Patterns
Chapter 23 Test Double Patterns
Chapter 24 Test Organization Patterns
Chapter 25 Database Patterns
Chapter 26 Design-for-Testability Patterns
Chapter 27 Value Patterns
PART IV: Appendixes
Appendix A Test Refactorings
Appendix B xUnit Terminology
Appendix C xUnit Family Members
Appendix D Tools
Appendix E Goals and Principles
Appendix F Smells, Aliases, and Causes
Appendix G Patterns, Aliases, and Variations
Gerard Meszaros is Chief Scientist and Senior Consultant at ClearStream Consulting, a Calgary-based consultancy specializing in agile development. He has more than a decade of experience with automated unit testing frameworks and is a leading expert in test automation patterns, refactoring of software and tests, and design for testability.
そういう人には、経験者を雇うか、「Working Effectively with Legacy Code」を読むよう著者は助言している。
The book consists of three parts. The first part is normal prose introducing the patterns and their contexts. The second part is a catalog of test smells and the last part (which is most of the book) describes the test patterns.
The first part of the book is excellent. It contains a couple of chapters related to test automation that are not specific to xUnit test automation. Chapter two introduces test smells, which I found very helpful. Chapter three contains the goals of test automation. This was one of the clearest descriptions I've seen which answer the question "why do we want to do this in the first place." The next chapters are about philosophy (e.g. TDD), principles and strategy. Chapter second introduces xUnit basics and the rest of the chapters of this part are narratives around the pattern sets (chapters) which are introduces further on in the book. The first part is about 180 pages of the book and is definitively worth reading completely. The individual chapters are short and easy to read.
The second part of the book covers test smells. The test smells are grouped intro three categories: code smells, behavior smells and project smells. These smells are also linked to each other, explaining what the "higher level smell" is caused by. I'm afraid I've seen all the smells in real life in the past and thus this part was very recognizable. Nowadays, I often send these smells to people I work with, since Gerard described them perhaps better than I ever could. The test smells are less than a hundred pages
The third part of the book are the test patterns. They are grouped in 'test pattern categories' with each category containing as little as four and as much as ten patterns. The pattern format Gerard uses is clear and easy to read. The largest drawback is that the pattern format causes quite a lot of duplication, especially across patterns. This is because the book is written so that each pattern can be read independently of others. This is good, but for people who read whole books (me) it causes the book to be somewhat boring to read (even though the content is interesting, just felt bored when reading again that there is one exception to object per test-case implementation of most xUnit frameworks).
The first chapter in part three is test strategy patterns, describing different approaches to testing (and this one is not specific to xUnit tests). The next chapter introduces xUnit frameworks written down in pattern form. Fixture setup is next in which the author describes the advantages and disadvantages for fresh vs shared fixtures and how to set them up. Chapter 21 then dives into the assertion parts of an xUnit framework describing the difference between state and behavior verification. Patterns such as custom assertions are frequently used in well-written test code. Next are the teardown patterns in which the authors somewhat promote the automated teardown (something I've not encountered very often in development). Chapter 23 is probably one of the best known chapters where Gerard categorizes test doubles (mocks/stubs/fakes) and describes when to use which term and why? This terminology is far from standardized and his effort to bring clarity in these terms is much appreciated. Next are test organization patterns such as how and where do you split your test case. Chapter 25 covers working with databases. Chapter 26 some general miscellaneous pattern in design which promote low coupling and thus increase testability. The final chapter relates to values within your tests.
The book also (for if this wasn't enough for you) contains over a hundred pages of appendixes. The refactorings, terminology mapping and glossary are useful.
Gerards book is huge. It contains useful insights related to unit testing and writing clean test code. From that perspective, this book is recommended for everyone serious about writing well-factored unit tests. However, the book also contains relative trivial things and, as mentioned, it contains a huge amount of duplication between different parts. This means that perhaps reading the book from cover to cover might not be the best approach :) Part one and two should definitively be read completely, but the patterns in chapter 3 are better browsed and read when needed or e.g. one per week. Another annoying part in the book, for me, was the way Gerard uses comments. Most comments in the code were not useful and made the code obscure and hard to read at times. Wished he had left them out (though, I guess this relates to coding style a lot).
Conclusion. Definitively worth reading or at least browsing for anyone serious about xUnit and test automation. Because of the duplication and the comments, I'd rate it four out of five stars. Personally I'd rather see a small 100 page book containing the summary and conclusions of what Gerard is talking about. It doesn't exist yet, so this book is, at this moment, the best on this particular subject.