Python Dependency Management

Speaker

Patrick Muehlbauer

Material

• video

Note

• Problems
  • multiple tools for the whole workflow
  • no real dependency resolution in pip
  • multiple files needed for reproducible env (e.g., requirements.txt, dev-requirements.txt)
• Possible Solutions
  • reproducible env
    • Pipfile, Pipfile.lock
  • distutils/setuptools limitations
    • pyproject.toml
• Python dependency managers
  • pipenv
  • pip-tools
  • poetry
  • hatch
  • conda
  • Bento
  • flit
  • pants/ pex
• Which to choose?
  • It depends.
  • Questions to ask
    • Are building a library or an application?
    • Doest it fit my infra?

pip-tools

• pip-compile: generates requirements.txt from setup.py or requirements.in
• pip-sync: dependency synchronization

Pipenv

• wrapper around pip and virtual environments
• use Pipfile, Pipfile.lock
• for applications rather than libraries
• separate sections for dev and man dependencies

Poetry

• tool to manage dependencies, build and publish packasge
• use pyproject.toml, poetry.lock

hatch

• simplifies development workflow
• warps multiple tools (pip, virtualenv, twine, pytest, coverage.py)
• setup.py, requirements.txt

Share on: Twitter ❄ Facebook ❄ Email

---