Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Investigating the role of use cases in the construction of class diagrams
Anda B., Sjøberg D. Empirical Software Engineering10 (3):285-309,2005.Type:Article
Date Reviewed: Jul 30 2007

Good software architecture is a mandatory prerequisite for a successful software project. How effective is the widely used use case analysis in helping establishing a solid software architecture and especially constructing a complete and well-structured class diagram? Anda and Sjøberg have investigated this problem by conducting two empirical experiments on how well use cases transform into class diagrams.

In the experiment setting, the authors have employed two alternative ways of using use cases. The first technique is called a derivation technique, where a domain model is created and then combined with use case-based scenarios to derive a class diagram. The second technique is called a validation technique, where an initial class diagram is constructed directly from the textual requirements and the use cases are used to validate and extend the initial class diagram.

These two techniques, derivation and validation, are then tested in two controlled experiments. The authors have done a great job designing and documenting the test setting and factors affecting the experiment. In the first experiment, 53 students were given a task of creating a class diagram using a textual requirements document and a use case model for a hypothetical library system. In the second experiment, the same task was given to 22 software professionals. In both cases, each group was divided into two: the first group used the derivation technique and the second the validation technique. The student group was further divided into “pen and paper” and “tool” groups, where the first group used only pen and paper and the other used a professional unified modeling language (UML) tool. All software professionals used only a UML tool of their choice. The resulting class diagrams were analyzed by independent software professionals for three qualities: completeness, structure, and the time required to create the class diagram.

I enjoyed the results. The students performed quite well compared to the professionals, and a UML tool did not contribute to class diagram quality as much as you might think. The authors show the results from the experiments in a table with statistical jargon, and sadly this makes the results more difficult to understand. The main result from these two experiments is that the derivation technique performs slightly better than the validation technique in the given experimental setting. The authors provide a good discussion of the validity of their experiment and the result they obtained.

This experiment feels like a model experiment from a scientific textbook for sociologists. The experiment is well designed and well thought out, and the results are systematically analyzed. A dark cloud in the experimental setting is the old debate about the validity of results obtained in a laboratory setting. Is it possible to do the same experiment in a real setting, using real problems with real developers in the middle of the everyday software circus, including panicking customers and manic bosses? I guess the answer is that we need both types of testing.

Reviewer:  Jari Ojala Review #: CR134578 (0807-0688)
Bookmark and Share
 
Object-Oriented Design Methods (D.2.2 ... )
 
 
State Diagrams (D.2.2 ... )
 
 
UML (D.3.2 ... )
 
 
Design Tools and Techniques (D.2.2 )
 
Would you recommend this review?
yes
no
Other reviews under "Object-Oriented Design Methods": Date
Understanding UML: the developer’s guide
Harmon P., Watson M., Morgan Kaufmann Publishers Inc., San Francisco, CA, 1998. Type: Book (9781558604650)
May 1 1999
Advanced object-oriented analysis and design using UML
Odell J., Cambridge University Press, New York, NY, 1998. Type: Book (9780521648196)
Oct 1 1998
Object modeling and design strategies
Gossain S., Cambridge University Press, New York, NY, 1998. Type: Book (9780521648226)
Oct 1 1998
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy