Social Network Analysis in Software Testing to Categorize Unit Test Cases Based on Coverage Information

Software testing is the most visible and cost-consuming activity in assuring the quality of software systems. In today's large-scale software systems, test (suite) maintenance is an inseparable part of software maintenance. Clarity of the purpose of each test case in the suite can be improved by proper naming and appropriate packaging, which decreases the cost of test maintenance. As a software system evolves its test suites need to be updated (maintained) to verify new or modified functionality of the software, and thus test cases may need to be categorized in a different way. In this work, we are proposing a technique to categorize the test cases automatically based on their coverage information. The proposed process can be performed dynamically over the life cycle of the system to improve the quality of test packaging. We build a social network of test cases and use coverage information to define links between them. This network is used to identify higher groups of test cases such as test package. To the best of our knowledge, this is the first trial in this direction. To evaluate our technique, we applied it on three open source systems with available JUnit test suits to identify test packages. We measured the quality of the discovered packages in terms of cohesion and coupling and compared them with the original packaging from test developers of these projects. The result shows that our technique can be used to categorize test cases automatically by even improving the quality of the packages.