Combining Data Mining Techniques for Evolutionary Analysis of Programming Languages

Abstract

Programming languages have been evolving gradually in response to changes in the programming industry. Many factors have been driving this evolution: for instance, improving language expressiveness, fixing bugs, and introducing new language features. However, modifying programming languages is a challenging process. One of the main difficulties is to gauge the perception of developers regarding the language over time. Thus, we set out to develop a framework aimed at evaluating the evolution of programming languages based on their technical documentation and the community's feedback from online discussions. Essentially, our framework is comprised of three main components: (1) Topic Modeling, which aims to extract the main semantic topics from the language aspects; (2) Sentiment Analysis, whose objective is to evaluate the perception of developers with respect to each identified topic; and (3) Data Visualization, which presents a visual metaphor that summarizes the information obtained in previous steps. To evaluate our proof-of-concept implementation of the framework, we carried out an evolutionary analysis of the Python programming language. According to our results, our framework was able to identify several changes made to the language as well as the programmers' perceptions regarding those changes: for instance, we found that the use of iterators over traditional repetition structures (i.e., count-based repetition) was initially received negatively by the community, but the outlook of developers on this new feature has matured enough for it to be considered beneficial to the programming language.