Feature-driven comparison of Git and Mercurial using DBSCAN and Random Forest
Year: 2025
Type: article
Abstract: Context: Git and Mercurial are two leading distributed version control systems as of today. Despite offering similar functionalities, they also differ in architectural designs that directly impact their performance and resource consumption.Objective: The work presented in this paper aims to systematically compare both tools using a machine learning-driven analysis to identify architectural disparities, focusing mainly on CPU usage, memory consumption, and repository size.Methods: The experiments were conducted on synthetically generated repositories, deliberately constructed to control and vary key performance-related features such as repository size, number of files, commits, and types of operations that are often inconsistent or unavailable in real-world repositories. To generate the results for comparison, DBSCAN was utilized to exclude performance anomalies, and Random Forest (RF) was used to analyze feature importance and determine their influence upon resource consumption.Results: Results indicate significant disparities, with Git’s compression mechanisms being more CPU-intensive, while Mercurial’s delta-based storage excels in memory efficiency. The analysis provides insights into the influence of repository characteristics as well as the operations performed on either tool on the consumption of resources.Conclusion: This research provides a data-driven evaluation of the trade-offs between these tools, offering insights for developers and organizations in selecting the most suitable DVCS based on specific project requirements.
Source: Array
Authors: Rashed Bahlool, Sameh Foulad, Sami Dagash
Institution: University of Bahrain
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,680
OpenAlex ID: W4414418457