Book Review: Object Oriented Analysis and Design with Applications, Third Edition

<Authors: Grady Booch, Robert A Maksimchuk, Michael W. Engle, Bobbi J. Young, Jim Conallen, Kelli A. Houston

Publisher: Addison-Wesley. Part of the Addison-Wesley Object Technology Series

ISBN: 978-0201895513 (Click to buy from Amazon)

This book covers how one should go about designing object oriented software, assuming no prior object oriented experience. It takes as its mission not only the purely technical aspects of that task, but also delves into requirements gathering, to ensure that the object oriented software delivered at the end is actually something the customer wants.

Some time ago, our company started work on a new software product. We had just begun designing the object model when I was given this book to review. This circumstance presented a great opportunity to not only read about object oriented design, but also to immediately apply the concepts, and see if those concepts actually worked in practice. I found that this book had three primary strengths.

The first strength is that when presenting a particular concept, the history of both that concept and the alternatives to that concept are presented. This allows the user to see why a particular concept is recommended over others, so that the user does not need to repeat the mistakes of past object oriented practitioners. For example, they trace the development of programming languages as they became more and more object oriented, showing why certain features were added to languages and the functionality that such features enabled, and how such features should be used.

The second strength is the advice the authors give in helping the user develop an object model. The authors quickly disabuse one of the notion that it easy to come up with a great object model on the first try. Instead, the authors state that developing a great object model is hard work, and the model will undergo transitions and become better as a project progresses. Specifically, they say “The best software designs look simple, but as experience shows, it takes a lot of hard work to design a simple architecture”. They continue by comparing the correct classification of logic and state into proper classes as part of a broader problem of classification in general that has been with us since Aristotle first classified plants and animals, and was continued through Aquinas, Descartes, and Locke. While this history of classification didn’t help me design an object model any faster, it at least made me feel better to know that it’s a problem that is known to be hard.

Even though object design is a hard problem, the authors do provide ways to help, by giving heuristics for development. They advocate an iterative approach to design, by gathering requirements and then gradually adding the requirements into the project. Doing so allows quick feedback as to the suitability of the model, allowing it to be easily changed.

The third strength is that in the latter half of the book, the authors attempt to apply the ideas given in the first part of the book to real projects. The value of this part of the book is not in looking at the resulting project, but rather in their explanations of how they attempt to correctly apply the ideas to the projects. For example, when building a web application, they use EJB 2.x, complete with the EJB interfaces, Facades, DTOs, and so on. While a new project started today would probably not make that technology choice, most of the time spent by the authors in this web application is not spent on these technology details, but rather on how the requirements were gathered, and how those requirements translated into the object model. Such information is useful no matter what the underlying technology.

One of the questions to be answered when reviewing a book for the Austin Java Users Group is “Is there an existing, canonical book that already covers the same ground.” This struck me as interesting, because this book itself is the canonical book for object oriented analysis and design. Its constant references to current literature (the bibliography is 73 pages long) help to assure the reader that he or she is actually using research proven methods to properly construct object models that will be useful, flexible, and correct. Recommended.