The effectiveness of partition testing

The effectiveness of partition testing

陳訓廷(Fun-ting Chan)

香港大學/Doctor of Philosophy/Computer Science/博士(1998年)

並列摘要

(Uncorrected OCR)
Abstract of the thesis entitled �he Effectiveness of Partition Testing�submitted by Fun Ting Chan for the degree of Doctor of Philosophy at The University of Hong Kong in May 1998.
Abstract
Partition testing is one of the commonly used testing strategies. The effectiveness of a partition testing strategy depends on its partitioning scheme and test case allocation scheme. The partitioning scheme governs how to divide the input domain into subdomains and the test case allocation scheme determines how to allocate test cases to each of the subdomains. Of course, the analysis of testing effectiveness is also related to the conditions of the testing environment and the metrics used in measuring the effectiveness. This research project extends the study of the effectiveness of partition testing along these four key areas: the test case allocation scheme, the partitioning scheme, the testing environment, and the testing effectiveness metric.
Regarding the test case allocation scheme, we compare, for the same partitioning scheme, the testing effectiveness of different test case allocation schemes. We identify several sufficient conditions for one test case allocation scheme to be more effective than another. We illustrate with case studies the applications of these findings for better testing effectiveness and discuss practical ways of applying them.
Regarding both the partitioning scheme and the test case allocation scheme, we introduce a cost effective partition testing strategy which is simple in concept and easy to apply. Its partitioning scheme simply divides the input domain under test into as many equal-sized subdomains as the number of test cases. Its test case
i
allocation scheme is to select one test case randomly from each subdomain. We refer to this strategy as the optimally refined proportional sampling strategy as it originates from the refinement of the proportional sampling strategy. However, the strategy can be used standalone and as a refinement to other partition testing strategies. This strategy incurs nominal partitioning overheads but gives better testing performance. Some guidelines to apply this strategy are given.
Regarding the testing environment, we extend the analytical study of testing effectiveness to cover the scenario of selecting test cases without replacement. This scenario is more practical but mathematically complex when compared with the normally used scenario of selecting test cases with replacement. We refer to these scenarios as the distinct scenario and the repetition scenario, respectively. We investigate the testing effectiveness of partition testing and random testing under the distinct scenario and compare the results with the repetition scenario.
Regarding the testing effectiveness metric, we introduce an improved metric that measures the number of distinct failures a test suite is capable of detecting. This new metric is a better indicator for the failure detection capability of a test suite when compared with the commonly used one in the literature. We examine the existing partition testing strategies in the light of this new metric. We find that the proportional sampling strategy and the optimally refined proportional sampling strategy remain to be favorable partition testing strategies over random testing.
ii

並列關鍵字

Partitions (Mathematics) ； Computational complexity.