How Do Contributors Impact Code Naturalness? An Exploratory Study of 50 Python Projects
Thanadon Bunkerd, Dong Wang, Raula Gaikovina Kula, Chaiyong Ragkhitwetsagul, Morakot Choetkiertikul, Thanwadee T. Sunetnanta, Takashi Ishio, Kenichi Matsumoto · 2019
Recent studies have shown how software is comparable to natural languages, meaning that source code is highly repetitive and predictable. Other studies have shown the naturalness as indicators for code quality (i.e., buggy code). With the rise of social coding and the popularity of open source projects, the software is now being built with contributions that come from contributors from diverse backgrounds. From this social contribution perspective, we explore how contributors impact code naturalness. In detail, our exploratory study investigators whether the developers' history of programming language experience affects the code naturalness. Calculating the code naturalness of 678 contributors from 50 open-source python projects, we analyze how two aspects of contributor activities impact the code naturalness: (a) the number of contributors in a software project, (b) diversity of programming language contributions. The results show that the code naturalness is affected by the diversity of contributors and that more collaborative software tends to be less predictable. This exploratory study serves as evidence into the relationship between code naturalness and the programming diversity of contributors.