Grant: Canada Research Chairs

When creating a software system, developers try to avoid re-inventing the wheel for every required functionality. Thus, they often rely on third-party software libraries that provide a toolbox of implementations and building blocks. However, not all libraries have the same quality, ease of use, or community support. Given many library alternatives to choose from, how can developers select the best library for their needs? What if they select a library but later find a better alternative, how can they automatically switch to the new library without having to re-write all the functionality already implemented using the old library? The goal of Dr. Nadi's research program is to provide automated support tools that help developers deal with such situations. Specifically, Dr. Nadi will develop metrics that help developers reason about important selection factors (e.g., library documentation quality and ease of use) . She will also design automated library migration techniques, especially for Python, that help developers easily switch between libraries without additional manual labor. The situations described above are ones that anyone developing a piece of code runs into. In fact, it is no longer only software developers in the traditional sense that face such situations when dealing with libraries; the new emerging group of data scientists that are now pervasive in most fields also face these same questions. However, given the diversity of background and training of data scientists (e.g., a social sciences or business background), navigating such situations may be even trickier than software developers. Thus, there may be additional factors and type of support to consider for data scientists and their highly specialized libraries. As the final direction of the proposed program, Dr. Nadi will investigate the needs of data scientists when it comes to library selection and usage, and customize the above support tools to these needs. Using innovative combinations of qualitative research, mining software repositories, machine learning, and program analysis, Dr. Nadi provides software library users with tools and techniques that help them spend less time on mundane activities. At the same time, it empowers them with data and explanations to help them accomplish more creative activities that need their expertise. Overall, such support improves both the efficiency and effectiveness with which complex code can be developed.