find_skrub_users

Search GitHub for public repositories that use the skrub package.

What it does

Runs a series of GitHub Code Search queries covering:

• Python import statements (import skrub, from skrub import ...)
• Dependency manifest files (requirements.txt, pyproject.toml, setup.py, setup.cfg, environment.yml, Pipfile)

Results are deduplicated by repository, sorted by stars, and written to both a CSV and a JSON file.

By default, forks of skrub-data/skrub itself are excluded.

Setup

python -m venv .venv
source .venv/bin/activate
pip install PyGithub

GitHub token

A GitHub personal access token is required (the search API requires authentication). You only need public access — no extra scopes are needed.

Create one at: https://github.com/settings/tokens

Either pass it with --token or set the GITHUB_TOKEN environment variable:

export GITHUB_TOKEN=ghp_xxxxxxxxxxxx

Usage

# Basic run — search all public repos
python find_skrub_users.py --token ghp_xxx

# Filter results to specific orgs or users
python find_skrub_users.py --token ghp_xxx --orgs inria,scikit-learn-contrib,probabl-ai

# Include forks of skrub-data/skrub (excluded by default)
python find_skrub_users.py --token ghp_xxx --include-skrub-forks

# Write output files to a specific directory
python find_skrub_users.py --token ghp_xxx --output-dir ./results

# Print queries without hitting the API
python find_skrub_users.py --dry-run

Output

Two files are written to --output-dir (default: current directory):

• skrub_users_YYYY-MM-DD.csv
• skrub_users_YYYY-MM-DD.json

Each row/entry represents one unique repository, with these fields:

Field	Description
repo_full_name	owner/repo
owner	GitHub username or org
stars	Star count at time of scan
url	Repository URL
last_pushed	ISO 8601 timestamp of last push
is_fork	Whether the repo is a fork of anything
is_skrub_fork	Whether it is specifically a fork of skrub-data/skrub
matched_files	Files that matched (semicolon-separated in CSV)
matched_queries	Which search queries found it (semicolon-separated in CSV)

Limitations

• The GitHub Code Search API caps results at 1,000 per query. If skrub usage grows significantly, some repos may not appear.
• Only the default branch of each repository is indexed.
• Only repositories active in the last year are indexed by GitHub.
• The search rate limit is 10 requests/minute — the script sleeps between queries automatically.
• Results reflect public repositories only.