Replication Package

Welcome to the replication package of the paper entitled "The Walking Packages: A Survival Analysis of ROS Repositories"

Repository Structure

• scripts: Complete data processing and analysis pipeline
  • 00-11: Core data collection pipeline (download ROS index, build mappings from rosdistro, extract repository features, apply exclusion criteria)
  • 12-14: Inflow analysis and documentation metrics
  • 15-17: Event tables and activity extraction (state machine, comments, reviews)
  • 18-21: Survival analysis pipeline (dataset preparation, survival models, extended analysis, paper figures)
  • Utility scripts: generate_all_commits_spreadsheet.py, generate_ros_packages_statistics.py
  • run_all.py: Execute scripts 00-11 sequentially
• out: Output directory for all results
  • survival_analysis: Survival analysis outputs (Kaplan-Meier plots, Cox model results, feature importance)
  • events: Repository event tables (commits, issues, PRs, comments, reviews)
  • repos: Per-repository data collected from GitHub (metadata, commits, contributors, issues, community files)
  • Key datasets: filtered_repo_dataset.csv,
• tables: Summarized statistics

Running the Replication Pipeline

To replicate our data collection and survival analysis:

pip install -r requirements.txt
echo "GITHUB_TOKEN=your_github_token_here" > .env
python scripts/run_all.py

The pipeline analyzes ROS packages across three distributions (ROS 2 Humble, Jazzy, and Kilted) and performs survival analysis following the methodology of Ait et al. (2022). The complete survival analysis dataset is generated at out/survival_dataset_complete.csv, with model results in out/survival_analysis/.

Individual scripts can be run independently for specific steps of the pipeline. Key scripts include:

• 09_extract_repo_features_and_commits.py: Extract repository metadata from GitHub (requires GitHub API token)
• 18_prepare_survival_dataset.py: Prepare survival analysis dataset with event definitions
• 19_survival_analysis.py: Run Kaplan-Meier, log-rank tests, and Cox proportional hazards models
• 21_paper_figures.py: Generate publication-ready figures

Contact

If you have any questions or are interested in contributing to this project, please don't hesitate to contact us:

• Juliana Freitas (jfreit4@lsu.edu)
• Elijah Phifer (ephife3@lsu.edu)
• Felipe Fronchetti (ffronchetti@lsu.edu)