To investigate whether a programming language is the right tool for the job, we gathered a very large data set from GitHub (728 projects, 63M lines of code, 29K authors, 1.5M commits, in 17 languages). Using a mixed-methods approach, combining multiple regression modeling with visualization and text analytics, we studied the effect of language features such as static v.s. dynamic typing, strong v.s. weak typing on software quality. By triangulating findings from different methods, and controlling for confounding effects such as code size, project age, and contributors, we observed that a language design choice does have a significant, but modest effect on software quality.
API stability and adoption in the Android Ecosystem
In today’s software ecosystem, which is primarily governed by web, cloud, and mobile technologies, APIs perform a key role to connect disparate software. Big players like Google, FaceBook, Microsoft aggressively publish new APIs to accommodate new feature requests, bugs fixes, and performance improvements. We investigated how such fast paced API evolution affects the overall software ecosystem? Our study on Android API evolution showed that the developers are hesitant to adopt fast evolving, unstable APIs. For instance, while Android updates 115 APIs per month on average, clients adopt the new APIs rather slowly, with a median lagging period of 16 months. Furthermore, client code with new APIs is typically more defect prone than the ones without API adaptation. To the best of my knowledge, this is the first work studying API adoption in a large software ecosystem, and the study suggests how to promote API adoption and how to facilitate growth of the overall ecosystems.
Assert Use in GitHub Projects
Assertions in a program are believed to improve software quality. We conducted a large scale study on how developers typically use assertions in C and C++ code and showed that they play positive role in improving code quality. We further characterized assertion usage along different process and product metrics. Such detailed characterization of assertions will help to predict relevant locations of useful assertions and eventually will improve code quality.