Adapting the returners and explorers dichotomy to understand source code knowledge distribution

Abstract

Understanding how knowledge about source code is distributed among contributors is critical to ensuring software quality and project sustainability over time. Established techniques, such as Truck Factor and authorship metrics, are based on contribution frequency and file ownership, but do not consider developers’ behavioral patterns when navigating the code base. This work adapts the Returners and Explorers (R&E) framework, originally proposed in the field of human mobility, to characterize developers’ interaction with source code based on orbital metrics. Using the concepts of code orbit radius (rco) and its variant located in the k most modified files (r(k) co ), developers can be classified as returners, explorers, or tourists, according to the dispersion of their modifications. The empirical analysis covers 20 open-source projects from the Apache Software Foundation, identifying that returners tend to concentrate on specific regions of the code, while explorers have a more comprehensive performance. Although often overlooked, tourists represent a significant portion of contributors, especially in peripheral files. A comparison between R&E profiles and expertise-based models, such as DOA, DOE, and Truck Factor, reveals that explorers often coincide with expert developers, while returners, although specialized, are sometimes underestimated. The R&E framework complements existing approaches by capturing behavioral dimensions of knowledge distribution, offering new perspectives on developer roles, collaboration patterns, and cognitive risks in large-scale software projects.