Is there a correlation between code comments and issues?: An exploratory study

Abstract

Comments in a software code base are one of the key artifacts that help developers in understanding the code with respect to development and maintenance. Comments provide us with the information that is used as a software metric to assess the code quality and which further can be applied to demonstrate its impact on the issues in the code. In this paper, we set out to understand the correlation between code comments and issues in Github. We conduct an empirical study on 625 repositories hosted on GitHub with Python as their primary language. We manually classify comments from a randomly selected sample of python repositories and then train and evaluate classifiers to automatically label comments as Relevant or Auxiliary. We extract the metadata of issues in each repository present in our dataset and perform various experiments to understand the correlation between code comments and issues. From our dataset of python repositories, we then plot a graph between the average time taken to resolve an issue and percentage of relevant comments in a repository to find if there is any relation or a pattern by which the latter affects the former. Our statistical approach of finding out the correlation between code comments and issues gives us the correlation factor by which code comments are related to issues. We conclude from our study that comments are indeed important and play an important role in solving issues of the project. We also found that increasing the percentage of relevant comments along with the source code can help in the reduction of the average number of days before an issue is resolved.