Equivalence class partitioning

  • Artigo

I have been teaching formal testing techniques for several years at Microsoft and University of Washington Extension. Techniques are systematic procedures to help solve a complex problem. A technique does not find all types problems; techniques are generally very good at finding very specific classes of defects. But, the usefulness or effectiveness of any particular technique relies on the in-depth system and domain knowledge of the tester, and in the tester's skill and ability to apply the correct technique to a problem space.

One of those techniques that I teach is equivalence class partitioning (ECP). On the surface ECP appears rather simple, but the effectiveness in the application of this technique relies on the tester's ability to decompose input and output variables into discrete subsets of valid and invalid classes.

Over-generalization of the data reduces the number of subsets in a specific class and increases the likelihood of under-testing. Hyper-decomposition of the variables may increase the number of subsets in a specific class without providing additional benefit (although it may potentially increase the number of tests). Based on this observation I have proposed the following hypothesis or Bj's Theory of Equivalence Class Partitioning Data Decomposition: Over-generalization of data reduces the baseline tests and increases the probability of missing errors or generating false positives. Hyper-analysis of data increases the probability of redundancy and reduces the overall effectiveness of each test.

ECP is not some rote activity, but in order for it to be most effective the tester requires a lot of in-depth knowledge about the overall system (hardware, operating system, environment, etc), the domain (client program, programming language used to develop the application, API interfaces, etc.), and the data sets. Limited knowledge or inability to adequately analyze the data based on the system and domain spaces in one or more of these areas is an impediment to the effective application of this technique. I suspect those who fail to perform an in-depth analysis of the system, domain, and data interaction leads some people disregard this technique, or become hyper-critical of its usefulness.

In this months issue of Software Test & Performance magazine I discuss the theory and application of the equivalence class partitioning technique as a viable and extremely useful tool used by professional testers. The magazine is free to download, and the article goes into great depth explaining when and how to apply the technique of equivalence class partitioning. Let me know if you have any questions, and I would appreciate your comments or feedback.

Comments

  • Anonymous
    October 01, 2007
    You might find some arguments and discussion about ECP at my blog .... As you can see lots of questions and less answers ... Would appereciate if you can I pick up some unanswered issues in my blog so that this post of yours can have some involved discussion ... (or I will come back after reading your STP article with my questions) http://shrinik.blogspot.com/2007/06/testers-quiz-what-is-equivalence.html Alan page dedicated one post on this topic on my request (when I met him at STAR EAST) http://blogs.msdn.com/alanpa/archive/2007/06/15/equivalence-class-partitioning.aspx this is an interesting discussion ... Thanks for starting this topic Shrini

  • Anonymous
    October 01, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    See my responses above in the comments

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    October 08, 2007
    The comment has been removed

  • Anonymous
    June 05, 2008
    I have been teaching formal testing techniques for several years at Microsoft and University of Washington Extension. Techniques are systematic procedures to help solve a complex problem. A technique does not find all types problems; techniques are generall