Technical Debt's State of Practice on Stack Overflow

Abstract

Context: Previous work have investigated the industry's state of practice on technical debt (TD) by surveying invited practitioners. However, these studies have included only practitioners with a sound understanding of TD, which might incur biases in their findings. One way to reduce this bias is to consider Q&A platforms, such as Stack Overflow (SO), because of containing practitioners with different roles and levels of TD knowledge. Therefore, the discussions on SO can broaden the perception of the state of practice on TD in the industry. Goal: To investigate how TD concepts have been discussed by SO's practitioners. Method: First, we employed data mining techniques to identify TD-related discussions on SO. Next, based on significance and popularity metrics, we considered 195 discussions to be further investigated. For each discussion, we analyzed its contents to identify types of TD followed by activities, strategies, and tools used in its management. Results: We found that code, infrastructure, and architecture debt are the most discussed types of TD on SO. Similarly, TD identification and payment are the most discussed activities. Regarding tools and strategies for TD management, SonarQube and refactoring have been mostly mentioned by practitioners. Conclusion: Our findings may indicate new research directions, as well as expanding the empirical knowledge on how the industry has handled TD.