软件测试中的全对偶测试理论
5星 · 超过95%的资源 需积分: 10 83 浏览量
2009-10-12
18:48:37
上传
评论
收藏 45KB PDF 举报 
Efficient Testing with All-Pairs
Prepared for STAREast 2003 International Conference on Software Testing
Bernie Berger
If you’re a software tester who’s been in the field for a few years, you may have found yourself in one of
the following situations:
• You’re working as hard as you can to find bugs in a huge system and you can’t get to everything
within the deadline. You’ve already stumbled across some good bugs, and you think there are
more in there, but the deadline comes and the software is released. A week later, a major client
finds a serious problem with the new release and lets everyone know about it in an industry press
release. You begin thinking about what went wrong and how you can improve your testing
coverage.
• You’re on a job interview, and the person across the desk asks you how to test a product,
especially when there’s too much to do in so little time.
• Your management has an elementary understanding of software testing, and as a result, sets
unrealistic expectations for you to “test everything.” They demand 100% coverage, including
testing all possible inputs, from all possible interfaces, into all possible system paths, into all
possible outputs. You know these are ridiculous demands, and you start thinking about alternate
testing methods. (and/or alternate employment opportunities).
Well, I’m able to cite these examples because I’ve been in each of these situations myself. A few years
ago, I started thinking about the Coverage question of testing software – how can you “test everything”
without really testing everything? How can you test efficiently: to minimize testing efforts but maximize
testing results?
I found a method that I enjoy so much that I use it and talk about it as often as possible. I’ve seen this
technique referred to as “Pairwise Testing,” “Combinatorial Method,” and “Orthogonal Arrays” (actually,
each of these is similar but different), but I’ll use the term “All-Pairs.”
In All-Pairs test design, we are concerned with variables of a system, and the possible values each variable
could take. We will generate testcases by pairing values of different variables. Don’t re-read that last
sentence -- generating testcases using All-Pairs is easier than it sounds. It’s like learning a new card game
– at first you have to learn the object of the game, all the rules, and all the exceptions to the rules, and then
the tips and strategies. But after you’ve played a few times, it seems naturally easy to play. It’s the same
here. The best way to explain how it works is with an example, so keep reading…
Acknowledgements: This presentation draws upon material from various sources,
especially Lessons Learned in Software Testing by Kaner, Bach and Pettichord. Many
thanks to them for their time and advice. Thanks to the AETG team who developed the
all-pairs approach. Thanks to the reviewers of this paper for their insightful feedback:
Lawrence
Nuanez, Michael Steinhart,
Dmitry
Shchelokov
资源评论
yingma_my2016-05-11英文版的，还没仔细看。- glacierrr2011-09-22本以为对偶论比较复杂的，这篇文章写的很简单易懂
