Achieving Quality through Software Maintenance and Evolution: on the role of Agile Methodologies and Open Source Software

Agile methodologies, open source software development, and emerging new technologies are at the base of disruptive changes in software engineering. Being effort estimation pivotal for effective project management in the agile context, in the first part of the thesis we contribute to improve effort estimation by devising a real-time story point classifier, designed with the collaboration of an industrial partner and by exploiting publicly available data on open source projects. We demonstrate that, after an initial training on at least 300 issue reports, the classifier estimates a new issue in less than 15 seconds with a mean magnitude of relative error between 0.16 and 0.61. In addition, issue type, summary, description, and related components prove to be project-dependent features pivotal for story point estimation. Since story points are the most popular effort estimation metric in the agile context, in the second study presented in the thesis we investigate the role of agile methodologies in software maintenance and evolution, and prove its undoubted influence on the refactoring research field over the last 15 years. In the later part of the thesis, we focus on recent technologies to understand their impact on software engineering. We start by proposing a specialized blockchain-oriented software engineering, on the basis of the peculiar challenges the blockchain sector must confront with and statistical data retrieved from a corpus of open source blockchain-oriented software repositories, identified relying upon the 2016 Moody’s Blockchain Report. We advocate the need for new professional roles, enhanced security and reliability, novel modeling languages, and specialized metrics, along with new research directions focusing on collaboration among large teams, testing, and specialized tools for the creation of smart contracts. Along with the blockchain, in the later part of this work we also study the growing mobile sector. More specifically, we focus on the relationships between software defects and the use of the underlying system API, proving that our findings are aligned with those in the literature, namely, that the applications which are more connected to API classes are also more defect-prone. Finally, in the last work presented in the dissertation, we conducted a statistical analysis of 20 open source object-oriented systems, 10 written in the highly popular language Java and 10 in the rising language Python. We leveraged two statistical distribution functions–the log-normal and the double Pareto distributions–to provide good fits, both in Java and Python, for three metrics, namely, the NOLM, NOM, and NOS metrics. The study, among other findings, revealed that the variability of the number of methods used in Python classes is lower than in Java classes, and that Java classes, on average, feature fewer lines of code than Python classes.