Why Virtual Environments Matter

Introduction

The Global Install Problem

# Without venv:
# pip install requests==2.25  -> global
# pip install requests==2.28  -> overwrites 2.25!
print('global conflict demo')

What is a Virtual Environment?

# venv creates:
# project/
#   .venv/
#     bin/python     (isolated Python)
#     lib/site-packages/  (project packages)
print('venv structure')

Reproducibility

# pip freeze > requirements.txt
# # Later, on another machine:
# python -m venv .venv
# pip install -r requirements.txt
print('reproducible env')

Security Isolation

# System Python is untouched:
# /usr/bin/python -> system packages
# .venv/bin/python -> project packages only
print('security isolation')

CI/CD and Docker

# Dockerfile pattern:
# RUN python -m venv /opt/venv
# RUN /opt/venv/bin/pip install -r requirements.txt
print('ci/cd demo')

One Venv Per Project

# Add to .gitignore:
# .venv/
# __pycache__/
print('gitignore tip')

System vs User vs venv Site-Packages

import sys
print('\n'.join(p for p in sys.path if 'site' in p))

pyenv for Python Version Management

# pyenv install 3.12.0
# pyenv local 3.12.0
# python -m venv .venv
print('pyenv demo')

conda as Alternative

# conda create -n myenv python=3.11
# conda activate myenv
# conda install numpy pandas
print('conda demo')

Best Practice Summary

import sys
print('Python:', sys.version)
print('venv:', sys.prefix != sys.base_prefix)

Quick Check

Recap

Keep Going