Browse Books

Reviewer: Andrew Brooks

Instructors now have a book on software testing that will allow them to move far beyond the simplistic black-box/white-box treatment that so often prevails in general software engineering textbooks. Those struggling to provide students with full and proper definitions of key terminologies and coverage criteria will find them all in the book's early chapters. Several coverage criteria are defined under each of four categories: graphs (structural and data flow coverage), logical expressions (predicate and clause coverage), input domain characterizations (input partitioning and combination coverage), and syntactic description (mainly various kinds of mutation coverage). Applications of the criteria are all well illustrated at various levels (specification, design, and coding), though fervent followers of unified modeling language (UML) might wish for more examples of activity diagrams, inheritance diagrams, state charts, and use cases. Figures 2.15, 3.1, and 4.2 show how the coverage criteria in each of the first three categories relate, including which criteria subsumes which. The treatment of subsumption is a particular strength of this book. In later chapters, the reader is provided with an understanding of the test process, and of the real and substantial difficulties of testing object-oriented, Web-based, real-time, and embedded software. Section 7.1 provides data flow criteria adapted for inheritance and polymorphism. A separate chapter covers how to approach the building of testing tools. Throughout the book, care is taken to distinguish material that is under active research, and not yet available in commercial tools. Each chapter closes with biographic notes that master's and doctoral students will find invaluable. The book's index, however, could be improved. The book does not discuss static analysis, usability testing, or practical JUnit usage. Bearing in mind that master's and doctoral students are among the book's target audiences, my criticisms of the book are as follows: three pages from the end, the authors admit that, despite many published experiments on testing, we still do not know enough to recommend a criteria to a development manager. A full chapter on experimentation should have been provided. Automatic nonrandom and random test data generation has become a reality. A full chapter on how such test data generation is achieved should have been provided. Static analysis techniques have matured considerably and are capable of detecting serious defects in software prior to dynamic testing. A full chapter on static analysis should have been provided. Last but not least, the oracle problem is the principal difficulty in testing modern, complex software. How do we know what the correct result should be__?__ Section 6.5, addressing this issue, amounts to a summary account, totaling only three pages. Again, a full chapter should have been provided. Despite the criticisms above, I strongly recommend this book to the software engineering community, especially instructors in software engineering programs. Put simply, it would be foolish to ignore so much useful material on software testing. Online Computing Reviews Service