'Whenever you read [Refactoring], it’s time to read it again. And if you haven’t read it yet, please do before writing another line of code.'
–David Heinemeier Hansson, Creator of Ruby on Rails, Founder & CTO at Basecamp
Fully Revised and Updated—Includes New Refactorings and Code Examples
“Any fool can write code that a computer can understand. Good programmers write code that humans can understand.”
–M. Fowler (1999)
For more than twenty years, experienced programmers worldwide have relied on Martin Fowler’s Refactoring to improve the design of existing code and to enhance software maintainability, as well as to make existing code easier to understand.
Like the original, this edition explains what refactoring is; why you should refactor; how to recognize code that needs refactoring; and how to actually do it successfully, no matter what language you use.
- Understand the process and general principles of refactoring
- Quickly apply useful refactorings to make a program easier to comprehend and change
- Recognize “bad smells” in code that signal opportunities to refactor
- Explore the refactorings, each with explanations, motivation, mechanics, and simple examples
- Build solid tests for your refactorings
- Recognize tradeoffs and obstacles to refactoring
Includes free access to the canonical web edition, with even more refactoring resources. (See inside the book for details about how to access the web edition.)
About the Author
Martin Fowler is Chief Scientist at ThoughtWorks. He describes himself as “an author, speaker, consultant and general loud-mouth on software development.” Fowler concentrates on designing enterprise software: exploring what makes a good design and what practices are needed to create one.
Table of Contents
Foreword to the First Edition xi
Chapter 1: Refactoring: A First Example 1
The Starting Point 1
Comments on the Starting Program 3
The First Step in Refactoring 5
Decomposing the statement Function 6
Status: Lots of Nested Functions 22
Splitting the Phases of Calculation and Formatting 24
Status: Separated into Two Files (and Phases) 31
Reorganizing the Calculations by Type 34
Status: Creating the Data with the Polymorphic Calculator 41
Final Thoughts 43
Chapter 2: Principles in Refactoring 45
Defining Refactoring 45
The Two Hats 46
Why Should We Refactor? 47
When Should We Refactor? 50
Problems with Refactoring 55
Refactoring, Architecture, and Yagni 62
Refactoring and the Wider Software Development Process 63
Refactoring and Performance 64
Where Did Refactoring Come From? 67
Automated Refactorings 68
Going Further 70
Chapter 3: Bad Smells in Code 71
Mysterious Name 72
Duplicated Code 72
Long Function 73
Long Parameter List 74
Global Data 74
Mutable Data 75
Divergent Change 76
Shotgun Surgery 76
Feature Envy 77
Data Clumps 78
Primitive Obsession 78
Repeated Switches 79
Lazy Element 80
Speculative Generality 80
Temporary Field 80
Message Chains 81
Middle Man 81
Insider Trading 82
Large Class 82
Alternative Classes with Different Interfaces 83
Data Class 83
Refused Bequest 83
Chapter 4: Building Tests 85
The Value of Self-Testing Code 85
Sample Code to Test 87
A First Test 90
Add Another Test 93
Modifying the Fixture 95
Probing the Boundaries 96
Much More Than This 99
Chapter 5: Introducing the Catalog 101
Format of the Refactorings 101
The Choice of Refactorings 102
Chapter 6: A First Set of Refactorings 105
Extract Function 106
Inline Function 115
Extract Variable 119
Inline Variable 123
Change Function Declaration 124
Encapsulate Variable 132
Rename Variable 137
Introduce Parameter Object 140
Combine Functions into Class 144
Combine Functions into Transform 149
Split Phase 154
Chapter 7: Encapsulation 161
Encapsulate Record 162
Encapsulate Collection 170
Replace Primitive with Object 174
Replace Temp with Query 178
Extract Class 182
Inline Class 186
Hide Delegate 189
Remove Middle Man 192
Substitute Algorithm 195
Chapter 8: Moving Features 197
Move Function 198
Move Field 207
Move Statements into Function 213
Move Statements to Callers 217
Replace Inline Code with Function Call 222
Slide Statements 223
Split Loop 227
Replace Loop with Pipeline 231
Remove Dead Code 237
Chapter 9: Organizing Data 239
Split Variable 240
Rename Field 244
Replace Derived Variable with Query 248
Change Reference to Value 252
Change Value to Reference 256
Chapter 10: Simplifying Conditional Logic 259
Decompose Conditional 260
Consolidate Conditional Expression 263
Replace Nested Conditional with Guard Clauses 266
Replace Conditional with Polymorphism 272
Introduce Special Case 289
Introduce Assertion 302
Chapter 11: Refactoring APIs 305
Separate Query from Modifier 306
Parameterize Function 310
Remove Flag Argument 314
Preserve Whole Object 319
Replace Parameter with Query 324
Replace Query with Parameter 327
Remove Setting Method 331
Replace Constructor with Factory Function 334
Replace Function with Command 337
Replace Command with Function 344
Chapter 12: Dealing with Inheritance 349
Pull Up Method 350
Pull Up Field 353
Pull Up Constructor Body 355
Push Down Method 359
Push Down Field 361
Replace Type Code with Subclasses 362
Remove Subclass 369
Extract Superclass 375
Collapse Hierarchy 380
Replace Subclass with Delegate 381
Replace Superclass with Delegate 399
Once upon a time, a consultant made a visit to a development project. The consultant looked at some of the code that had been written; there was a class hierarchy at the center of the system. As he wandered through the hierarchy, the consultant saw that it was rather messy. The higher-level classes made certain assumptions about how the classes would work, assumptions that were embodied in inherited code. That code didn't suit all the subclasses, however, and was overridden quite heavily. If the superclass had been modified a little, then much less overriding would have been necessary. In other places some of the intention of the superclass had not been properly understood, and behavior present in the superclass was duplicated. In yet other places several subclasses did the same thing with code that could clearly be moved up the hierarchy.
The consultant recommended to the project management that the code be looked at and cleaned up, but the project management didn't seem enthusiastic. The code seemed to work and there were considerable schedule pressures. The managers said they would get around to it at some later point.
The consultant had also shown the programmers who had worked on the hierarchy what was going on. The programmers were keen and saw the problem. They knew that it wasn't really their fault; sometimes a new pair of eyes are needed to spot the problem. So the programmers spent a day or two cleaning up the hierarchy. When they were finished, the programmers had removed half the code in the hierarchy without reducing its functionality. They were pleased with the result and found that it became quicker and easier both to add new classes to the hierarchy and to use the classes in the rest of the system.
The project management was not pleased. Schedules were tight and there was a lot of work to do. These two programmers had spent two days doing work that had done nothing to add the many features the system had to deliver in a few months time. The old code had worked just fine. So the design was a bit more 'pure' a bit more 'clean.' The project had to ship code that worked, not code that would please an academic. The consultant suggested that this cleaning up be done on other central parts of the system. Such an activity might halt the project for a week or two. All this activity was devoted to making the code look better, not to making it do anything that it didn't already do.
How do you feel about this story? Do you think the consultant was right to suggest further clean up? Or do you follow that old engineering adage, 'if it works, don't fix it'?
I must admit to some bias here. I was that consultant. Six months later the project failed, in large part because the code was too complex to debug or to tune to acceptable performance.
The consultant Kent Beck was brought in to restart the project, an exercise that involved rewriting almost the whole system from scratch. He did several things differently, but one of the most important was to insist on continuous cleaning up of the code using refactoring. The success of this project, and role refactoring played in this success, is what inspired me to write this book, so that I could pass on the knowledge that Kent and others have learned in using refactoring to improve the quality of software.What Is Refactoring?
Refactoring is the process of changing a software system in such a way that it does not alter the external behavior of the code yet improves its internal structure. It is a disciplined way to clean up code that minimizes the chances of introducing bugs. In essence when you refactor you are improving the design of the code after it has been written.
'Improving the design after it has been written.' That's an odd turn of phrase. In our current understanding of software development we believe that we design and then we code. A good design comes first, and the coding comes second. Over time the code will be modified, and the integrity of the system, its structure according to that design, gradually fades. The code slowly sinks from engineering to hacking.
Refactoring is the opposite of this practice. With refactoring you can take a bad design, chaos even, and rework it into well-designed code. Each step is simple, even simplistic. You move a field from one class to another, pull some code out of a method to make into its own method, and push some code up or down a hierarchy. Yet the cumulative effect of these small changes can radically improve the design. It is the exact reverse of the normal notion of software decay.
With refactoring you find the balance of work changes. You find that design, rather than occurring all up front, occurs continuously during development. You learn from building the system how to improve the design. The resulting interaction leads to a program with a design that stays good as development continues.What's in This Book?
This book is a guide to refactoring; it is written for a professional programmer. My aim is to show you how to do refactoring in a controlled and efficient manner. You will learn to refactor in such a way that you don't introduce bugs into the code but instead methodically improve the structure.
It's traditional to start books with an introduction. Although I agree with that principle, I don't find it easy to introduce refactoring with a generalized discussion or definitions. So I start with an example. Chapter 1 takes a small program with some common design flaws and refactors it into a more acceptable object-oriented program. Along the way we see both the process of refactoring and the application of several useful refactorings. This is the key chapter to read if you want to understand what refactoring really is about.
In Chapter 2 I cover more of the general principles of refactoring, some definitions, and the reasons for doing refactoring. I outline some of the problems with refactoring. In Chapter 3 Kent Beck helps me describe how to find bad smells in code and how to clean them up with refactorings. Testing plays a very important role in refactoring, so Chapter 4 describes how to build tests into code with a simple open-source Java testing framework.
The heart of the book, the catalog of refactorings, stretches from Chapter 5 through Chapter 12. This is by no means a comprehensive catalog. It is the beginning of such a catalog. It includes the refactorings that I have written down so far in my work in this field. When I want to do something, such as Replace Conditional with Polymorphism (255), the catalog reminds me how to do it in a safe, step-by-step manner. I hope this is the section of the book you'll come back to often.
In this book I describe the fruit of a lot of research done by others. The last chapters are guest chapters by some of these people. Chapter 13 is by Bill Opdyke, who describes the issues he has come across in adopting refactoring in commercial development. Chapter 14 is by Don Roberts and John Brant, who describe the true future of refactoring, automated tools. I've left the final word, Chapter 15, to the master of the art, Kent Beck.Refactoring in Java
For all of this book I use examples in Java. Refactoring can, of course, be done with other languages, and I hope this book will be useful to those working with other languages. However, I felt it would be best to focus this book on Java because it is the language I know best. I have added occasional notes for refactoring in other languages, but I hope other people will build on this foundation with books aimed at specific languages.
To help communicate the ideas best, I have not used particularly complex areas of the Java language. So I've shied away from using inner classes, reflection, threads, and many other of Java's more powerful features. This is because I want to focus on the core refactorings as clearly as I can.
I should emphasize that these refactorings are not done with concurrent or distributed programming in mind. Those topics introduce additional concerns that are beyond the scope of this book.Who Should Read This Book?
This book is aimed at a professional programmer, someone who writes software for a living. The examples and discussion include a lot of code to read and understand. The examples are all in Java. I chose Java because it is an increasingly well-known language that can be easily understood by anyone with a background in C. It is also an object-oriented language, and object-oriented mechanisms are a great help in refactoring.
Although it is focused on the code, refactoring has a large impact on the design of system. It is vital for senior designers and architects to understand the principles of refactoring and to use them in their projects. Refactoring is best introduced by a respected and experienced developer. Such a developer can best understand the principles behind refactoring and adapt those principles to the specific workplace. This is particularly true when you are using a language other than Java, because you have to adapt the examples I've given to other languages.
Here's how to get the most from this book without reading all of it.
- If you want to understand what refactoring is, read Chapter 1; the example should make the process clear.
- If you want to understand why you should refactor, read the first two chapters. They will tell you what refactoring is and why you should do it.
- If you want to find where you should refactor, read Chapter 3. It tells you the signs that suggest the need for refactoring.
- If you want to actually do refactoring, read the first four chapters completely. Then skip-read the catalog. Read enough of the catalog to know roughly what is in there. You don't have to understand all the details. When you actually need to carry out a refactoring, read the refactoring in detail and use it to help you. The catalog is a reference section, so you probably won't want to read it in one go. You should also read the guest chapters, especially Chapter 15.
I need to say right now, at the beginning, that I owe a big debt with this book, a debt to those whose work over the last decade has developed the field of refactoring. Ideally one of them should have written this book, but I ended up being the one with the time and energy.
Two of the leading proponents of refactoring are Ward Cunningham and Kent Beck. They used it as a central part of their development process in the early days and have adapted their development processes to take advantage of it. In particular it was my collaboration with Kent that really showed me the importance of refactoring, an inspiration that led directly to this book.
Ralph Johnson leads a group at the University of Illinois at Urbana-Champaign that is notable for its practical contributions to object technology. Ralph has long been a champion of refactoring, and several of his students have worked on the topic. Bill Opdyke developed the first detailed written work on refactoring in his doctoral thesis. John Brant and Don Roberts have gone beyond writing words into writing a tool, the Refactoring Browser, for refactoring Smalltalk programs.Acknowledgments
Even with all that research to draw on, I still needed a lot of help to write this book. First and foremost, Kent Beck was a huge help. The first seeds were planted in a bar in Detroit when Kent told me about a paper he was writing for the Smalltalk Report Beck, hanoi. It not only provided many ideas for me to steal for Chapter 1 but also started me off in taking notes of refactorings. Kent helped in other places too. He came up with the idea of code smells, encouraged me at various sticky points, and generally worked with me to make this book work. I can't help thinking he could have written this book much better himself, but I had the time and can only hope I did the subject justice.
As I've written this, I wanted to share much of this expertise directly with you, so I'm very grateful that many of these people have spent some time adding some material to this book. Kent Beck, John Brant, William Opdyke, and Don Roberts have all written or co-written chapters. In addition, Rich Garzaniti and Ron Jeffries have added useful sidebars.
Any author will tell you that technical reviewers do a great deal to help in a book like this. As usual, Carter Shanklin and his team at Addison-Wesley put together a great panel of hard-nosed reviewers. These were
- Ken Auer, Rolemodel Software, Inc.
- Joshua Bloch, Javasoft
- John Brant, University of Illinois at Urbana-Champaign
- Scott Corley, High Voltage Software, Inc.
- Ward Cunningham, Cunningham & Cunningham, Inc.
- Stephane Ducasse
- Erich Gamma, Object Technology International, Inc.
- Ron Jeffries
- Ralph Johnson, University of Illinois
- Joshua Kerievsky, Industrial Logic, Inc.
- Doug Lea, SUNY Oswego
- Sander Tichelaar
They all added a great deal to the readability and accuracy of this book, and removed at least some of the errors that can lurk in any manuscript. I'd like to highlight a couple of very visible suggestions that made a difference to the look of the book. Ward and Ron got me to do Chapter 1 in the side-by-side style. Joshua suggested the idea of the code sketches in the catalog.
In addition to the official review panel there were many unofficial reviewers. These people looked at the manuscript or the work in progress on my Web pages and made helpful comments. They include Leif Bennett, Michael Feathers, Michael Finney, Neil Galarneau, Hisham Ghazouli, Tony Gould, John Isner, Brian Marick, Ralf Reissing, John Salt, Mark Swanson, Dave Thomas, and Don Wells. I'm sure there are others who I've forgotton; I apologize and offer my thanks.
A particularly entertaining review group is the infamous reading group at the University of Illinois at Urbana-Champaign. Because this book reflects so much of their work, I'm particularly grateful for their efforts captured in real audio. This group includes Fredrico 'Fred' Balaguer, John Brant, Ian Chai, Brian Foote, Alejandra Garrido, Zhijiang 'John' Han, Peter Hatch, Ralph Johnson, Songyu 'Raymond' Lu, Dragos-Anton Manolescu, Hiroaki Nakamura, James Overturf, Don Roberts, Chieko Shirai, Les Tyrell, and Joe Yoder.
Any good idea needs to be tested in a serious production system. I saw refactoring have a huge effect on the Chrysler Comprehensive Compensation system (C3). I want to thank all the members of that team: Ann Anderson, Ed Anderi, Ralph Beattie, Kent Beck, David Bryant, Bob Coe, Marie DeArment, Margaret Fronczak, Rich Garzaniti, Dennis Gore, Brian Hacker, Chet Hendrickson, Ron Jeffries, Doug Joppie, David Kim, Paul Kowalsky, Debbie Mueller, Tom Murasky, Richard Nutter, Adrian Pantea, Matt Saigeon, Don Thomas, and Don Wells. Working with them cemented the principles and benefits of refactoring into me on a firsthand basis. Watching their progress as they use refactoring heavily helps me see what refactoring can do when applied to a large project over many years.
Again I had the help of J. Carter Shanklin at Addison-Wesley and his team: Krysia Bebick, Susan Cestone, Chuck Dutton, Kristin Erickson, John Fuller, Christopher Guzikowski, Simone Payment, and Genevieve Rajewski. Working with a good publisher is a pleasure; they provided a lot of support and help.
Talking of support, the biggest sufferer from a book is always the closest to the author, in this case my (now) wife Cindy. Thanks for loving me even when I was hidden in the study. As much time as I put into this book, I never stopped being distracted by thinking of you.