Investigating the effects of design quality in the OSS projects
Software systems are becoming an essential part of the lives of both individuals and organizations, and as a consequence, these systems are getting bigger and more complex. Because of this, the tasks of maintaining the quality in these complex software systems are becoming increasingly difficult. Furthermore, these systems are subject to constant pressure to add modifications, implementing new features, or regular bug fixing. Due to an often-strict release schedule, developers may not get the chance to design and implement ideal solutions. This can lead to decay in software design and the growth of technical debt, which negatively impacts the overall quality of the software. Design quality plays a vital role in maintaining the quality of software systems. In the past, several studies have shown the impact of design quality on different quality attributes such as bug density, productivity, and maintenance. Researchers also proposed several tools and techniques to improve design quality. Despite efforts by the research community in past years, there are still gaps in our understanding of how does design quality evolves in OSS projects, how projects manage design quality and how does design quality impacts the long-term health of the projects. This thesis proposes a number of large-scale empirical investigations aiming to understand the gaps. In the first part of this thesis, we investigate the evolution of design quality in OSS projects. Researchers used different metrics to measure the project’s design quality. In this thesis, we investigated two different design quality metrics. The first one is code smells, which are the result of poor design or implementation choices. And the second one is entropy, provides an understanding of design quality in terms of flexibility during the enhancement of its functionality. Our results show that design issues build up and design quality (measured by code smells and entropy) degrades over time. As the project grows older and larger, design issues are fixed less and, as a consequence, build up. In the second part, we investigate how OSS projects manage design quality. As design discussion is the primary mechanism for OSS projects to debate, make and document design decisions, we investigated how developers discuss design in the community, how they evolve along with design quality, and what effect they have on the design quality. Our results show that: I) Regardless of different communication channels used for discussion, 89:51% of all design discussions occur in the project mailing list, II) the average number of design discussions decrease; design quality degrade, as the project evolves, and III) the correlation between design discussions and design quality is small. In the third part, we investigate how does design quality impacts day to day activity (such as code merging) as well as long term health of the projects. We find that code with design issues (code smells) are more likely to be involved in merge conflicts. Our results also show that both code smells and entropy have an impact on the long-term health of the project. In this thesis, we use bug-proneness of the codebase as the measurement of the project’s long-term health.
Co Advisor: Anita Sarma
Co Advisor: Carlos Jensen
Committee: Margaret M. Burnett
Committee: Arash Termehchy
GCR: Margaret Louise Niess
Monday, August 31 at 10:00am to 12:00pmVirtual Event