Software Development Patterns
This time I look at three books about programming. None of them has the word "patterns" in its title, and none of them claims to deal with patterns or best practices, but that is what they are all about.
Effective Java, 2d ed by Joshua Block (Addison-Wesley, Upper Saddle River NJ, 2008, 368pp, ISBN 978-0-321-35668-0, $49.99)
When I reviewed the first edition of this book (Micro Review, July/Aug 2002), I had never met Josh Bloch. Several years later, however, I was visiting the Google booth at JavaOne, when Josh came up to me and told me that my review had launched his book. He believed this because when my review appeared, his book had already been out for a year with modest sales, and his sales then increased substantially. I don't believe that my review had much to do with it. The book has a lot of substance, and it takes time for people to digest it and spread the word. Nonetheless, I recently took advantage of his gratitude to persude him to give me a copy of the new edition.
The first edition contains 57 short essays, called items. The new edition has 78 items. A few of the original items have disappeared, and Josh has added items to reflect the important changes to Java since 2001. He was a major player on the teams that developed generics, enum types, annotations, autoboxing, and the Java 5 for-each loop and concurrency library.
In 2004, Josh moved from Sun to Google, where he continues to guide the evolution of the language but now also develops Java libraries that use the new features. Thus he knows Java both as a language designer and a language user. There is probably no other person who could have written this book with the same degree of insight and experience.
If you ever have a chance to hear Josh talk, don't miss it. He is an excellent teacher, because he explains complex points clearly. That talent shines through the pages of this book as well. The items that Josh discusses are quite complex, even abstruse. Yet any experienced Java programmer can understand them.
My review of the first edition highlights many good points that are still good in this one. What I concluded about the first edition remains true of this one. Every bit of this book is essential for Java designers. Reading this book before you start delivering products can easily repay its cost thousands of times.
The ThoughtWorks Anthology -- Essays on Software Technology and Innovation by members of the ThoughtWorks staff (Pragmatic Bookshelf, Raleigh NC, 2008, 234pp, ISBN 978-1-934356-14-2, $38.95)
ThoughtWorks is a software development firm that specializes in fighting software bloat. To this end they have developed techniques for programming, building, testing, and delivering applications efficiently. The essays in this book present several of these techniques.
ThoughtWorks's founder, Roy Singham lays out his vision in the first essay. He talks about "the last mile" of putting a system into production. The expression "the last mile" comes from the telecommunications industry. The high speed backbone of the global internet slows down when it reaches the "last mile" of copper wire that brings bits to and from most individual customers. People have been talking about this problem for decades, but customers have seen few improvements.
For Singham, the last mile starts when the software meets all of its functional requirements. It ends when it is in place, delivering value to the customer. Development teams can bring a project to a conclusion rapidly and efficiently because of their tools and methodologies. Customers then begin a long, slow, expensive process. They must ensure that the new software can handle the expected workload, work well with other systems, interact with existing data without corrupting it, and protect against all sorts of unauthorized use or attacks.
Currently, developers derive functional requirements from end users who, ideally, have continual contact with the developers throughout the project. Singham suggests adding the last mile tasks to the functional requirements and involving the last mile stakeholders in the project from its inception. This should lead developers to automate much of the last mile process, just as they routinely automate builds and testing.
In telecommunications, the Holy Grail of the last mile problem is a means by which the internet can reach any customer at full speed (for example, fiber to the home). For software, Singham sees the Holy Grail as versionless software, a means by which developers can add features directly to production systems "on the fly," without fear of unintended consequences. Neither of these visions will come to pass any time soon.
Twelve more essays round out this book. They deal with programming, managing development, maintaining web applications and service oriented architectures, and building and testing systems. Having worked with large, convoluted Ant scripts, I especially appreciate Julian Simpson's essay on refactoring techniques for such files.
In the 1970s I worked with the late Rudolph Langer, who built a number of special purpose languages, usually on top of macro assembly languages. One of these, for example, facilitated designing pinball machines. Another made it easy to design optical forms to enable users to enter data visually. Nowadays we might call these domain specific languages (DSLs). Martin Fowler's essay explains how to use Ruby to create a DSL.
One of my favorite essays is Jeff Bay's "Object Calisthenics." He proposes an exercise in which you must write a 1000 line Java program using the following extremely strict coding standards:
- Use only one level of indentation per method.
- Don't use the else keyword.
- Wrap all primitives and strings.
- Use only one dot per statement.
- Don't abbreviate.
- Keep all entities small.
- Limit classes to two instance variables.
- Use first class collections.
- Don't use getters, setters, or properties.
Bay explains the points that he intends this exercise to make. He wants you to encapsulate data, use polymorphism properly, avoid duplication, and appropriately model the problem that you want the code to solve.
Like most books from the Pragmatic Bookshelf, this one is full of good practical advice. If you develop software, this one should be on your bookshelf.
Cockford's explanations of how to avoid the bad parts entail sophisticated programming. His examples require careful reading, and real mastery probably takes a lot of practice. Cockford characterizes the book as follows:
"This book is small, but it is dense. There is a lot of material packed into it. Don't be discouraged if it takes multiple readings to get it."