On the interplay between developer knowledge diversity and test code quality: understanding and addressing test smells in software development.

Abstract

Software development is a collaborative, social, knowledge-intensive activity, and human-centered aspects such as communication and personality can impact software projects. These factors are essential for team diversity. There are three types of team diversity: informational (or knowledge), social, and values diversity. In this context, we have noticed a lot of interest in the software engineering community about the relationship between human factors and code quality. Investigating if knowledge diversity (human aspects) affects test code quality is important. The literature has brought up several studies investigating test code quality. Still, little empirical evidence exists on the effects of knowledge diversity (e.g., educational background, level of professional experience, expertise, and skills) and test code quality. Whether we consider the prevalence of test smells in current software testing research, the number is way more limited. Test smells are bad implementations inserted by developers and can harm the comprehensibility and maintainability of test suites. Recent studies discuss developers’ perceptions of test smells and their impact on quality improvement, but there is little evidence regarding the effects of knowledge diversity in this direction, especially concerning interdisciplinary collaboration and the exchange of experiences between teams. This thesis aims to gather empirical evidence on the relationship between developer knowledge diversity and test code quality, particularly the effect test smells can bring on software quality. Initially, we built a knowledge base by considering the main concepts of software testing, maintenance, evolution, test smells, software refactoring, developer classification, and tool support. Next, to accomplish our research objective, we used a mixed-methods approach (e.g., surveys, interviews, and mining of GitHub repositories). We organized the findings as a set of guidelines to support developers in preventing the insertion of test smells during the creation of unit test cases.