Bug analysis in Jupyter notebook projects: an empirical study.

Abstract

One of the new technologies driving data science projects is Computational Notebooks, which allow users to build data-oriented codes, emphasizing the data collected and the analysis performed. Although Computational Notebooks have gained visibility, some problems and solutions already discussed and studied by the software engineering community must be addressed, impacting the quality of the developed software and, consequently, data analysis. In addition, neglecting these aspects can lead to the spread of bad programming practices. Computational Notebooks, such as Jupyter, have been widely adopted by data scientists to write code for analyzing and visualizing data. Despite their growing adoption and popularity, few studies are available to understand Jupyter development challenges from the practitioners’ point of view. This dissertation systematically studies bugs and challenges that Jupyter practitioners face through a large-scale empirical investigation. We mined 14,740 commits from 105 GitHub open-source projects with Jupyter Notebook code. Next, we analyzed 30,416 StackOverflow posts, which gave us insights into bugs that practitioners face when developing Jupyter Notebook projects. We conducted nineteen interviews with data scientists to uncover more details about Jupyter bugs and to gain insight into Jupyter developers’ challenges and finally, to validate all the information obtained, we carried out a survey with several data scientists and an analysis with association rules using the Apriori algorithm. We propose a bug taxonomy for Jupyter projects based on our results. We also highlight bug categories, their root causes, and Jupyter practitioners’ challenges.