Git & GitHub for Data Science

Version Control for Analysis Projects

Your Name

Why Version Control?

Real scenario:

“I broke my analysis script and need to go back to yesterday’s version”

  • Track changes over time
  • Collaborate without chaos
  • Experiment safely with branches

What is Git?

  • Version control system - tracks changes in your files
  • Works locally on your computer
  • Creates snapshots (commits) of your work
  • Like “track changes” for code and data projects

What is GitHub?

  • Remote hosting for Git repositories
  • Backup in the cloud
  • Share and collaborate
  • Portfolio of your work

Key distinction: Git = local, GitHub = remote

Git Basics

What Git Tracks

✅ DO track:

  • Code (.py, .R files)
  • Notebooks (.ipynb, .qmd)
  • Documentation (README, notes)
  • Configuration files

❌ DON’T track:

  • Large datasets
  • API keys/credentials
  • Generated outputs

Core Concepts

Repository (repo): Project folder Git tracks

Commit: Snapshot of your work at a point in time

Branch: Separate line of development (we’ll keep it simple today)

Basic Workflow

1. Modify files
2. Stage changes (git add)
3. Commit changes (git commit)
4. Repeat!

Essential Commands

The 5 Commands You Need

# Start tracking a project
git init

# Stage changes
git add filename.py

# Save a snapshot
git commit -m "Clear message about what changed"

# Check status
git status

# View history
git log

Live Demo

Creating a simple analysis repository

git init my-analysis
cd my-analysis
# Create a simple notebook or script
git add analysis.ipynb
git commit -m "Initial analysis of dataset X"

GitHub for Collaboration

Connecting to GitHub

Two new commands:

# Send commits to GitHub
git push

# Get latest changes from GitHub
git pull

Typical Workflow

  1. Create repo on GitHub
  2. Connect local repo to GitHub
  3. Push your work
  4. Collaborate with others
  5. Pull their changes

Demo: Push to GitHub

# Connect to GitHub (one time)
git remote add origin https://github.com/username/repo.git

# Push your commits
git push -u origin main

Now anyone can see and use your analysis!

Data Science Specifics

What to Commit

Essential files:

  • Analysis scripts (.py, .R)
  • Jupyter/Quarto notebooks
  • requirements.txt or renv.lock
  • README.md with project description
  • Documentation

What NOT to Commit

Use .gitignore for:

  • Data files (*.csv, *.parquet) - usually too large
  • Credentials (.env, API keys)
  • Output files (plots, HTML reports)
  • Virtual environments (venv/, renv/)
  • Jupyter checkpoints (.ipynb_checkpoints/)

The .gitignore File

Create .gitignore in your repo:

# Data
*.csv
*.parquet
data/

# Credentials
.env
secrets.json

# Outputs
*.html
figures/

# Python
__pycache__/
venv/

Jupyter Notebooks

Challenge: Notebooks contain outputs and metadata

Tips:

  • Clear outputs before committing (Cell → All Output → Clear)
  • Use tools like nbstripout to automate
  • Or use Quarto instead (.qmd files are cleaner!)

Hands-On Exercise

Your Task (10 minutes)

  1. Create a new directory for a simple analysis
  2. Initialize Git (git init)
  3. Create a basic notebook or script
  4. Make your first commit
  5. Make a change and commit again
  6. Create a GitHub repo and push

Bonus: Add a .gitignore file

Wrap Up

Key Takeaway

Commit early, commit often, write clear messages

Good commit message: "Add data cleaning function for missing values"

Bad commit message: "updates" or "fixed stuff"

Resources

Questions?

Remember: Everyone struggles with Git at first. It gets easier with practice!