Python Dependency Management - PyconDE 2018
- 1. Python Dependency Management Patrick Muehlbauer (@tmuxbee) 2018-10-24
- 2. • Predicting bananas • ...with Python Blue Yonder
- 5. How to provide this to others?
- 6. $ python setup.py sdist banana-predictions-1.0.0.tar.gz
- 7. $ python setup.py bdist_wheel banana_predictions-1.0.0-py3-none-any.whl
- 8. How to upload to PyPI?
- 9. $ python setup.py upload but no HTTPS
- 10. twine
- 11. How to specify dependencies?
- 13. How to install other uploaded packages?
- 14. pip
- 15. Different versions of ‘banana_predictions’ in different projects?
- 16. virtualenv $ virtualenv –p python3.7 venv
- 18. $ pip freeze > requirements.txt $ pip install –r requirements.txt
- 19. • multiple tools for the whole workflow • no real dependency resolution in pip • multiple files needed for reproducible environments (requirements.txt, dev-requirements.txt, …)
- 23. pyproject.toml PEP 517 -- A build-system independent format for source trees PEP 518 -- Specifying Minimum Build System Requirements for Python Projects
- 27. pip-tools • set of command line tools: pip-compile and pip-sync • pip-compile for dependency resolution: generates requirements.txt from setup.py or a requirements.in • pip-sync for dependency synchronization
- 31. pip-tools brings dependency resolution and a nicer pinning workflow
- 32. Pipenv • wrapper around pip and virtual environments • lots of convenience features (e.g. automatic virtualenv creation, python installation) • uses Pipfile and Pipfile.lock in addtion to setup.py • main focus on applications rather than libraries
- 33. Pipfile • separate sections for dev and main dependencies instead of requirements.txt and dev- requirements.txt • optional section to define required Python version
- 34. Pipfile.lock
- 35. $ pipenv install scipy $ pipenv uninstall scipy update Pipfile and Pipfile.lock
- 36. Poetry • tool to manage dependencies, build and publish packages • automatic virtualenv creation • single config file for everything: pyproject.toml, poetry.lock for reproduciblity • reliable dependency resolution • project scaffolding
- 37. $ poetry new my-project
- 38. $ poetry add scipy
- 39. $ poetry version
- 40. $ poetry build sdist and .whl
- 41. $ poetry publish
- 42. poetry.lock
- 43. hatch • simplifies development workflow • wraps multiple tools (pip, virtualenv, twine, pytest, coverage.py, …) • setup.py + requirements.txt
- 44. $ hatch new my-project
- 45. $ hatch install
- 46. $ hatch test
- 47. $ hatch grow
- 48. $ hatch build sdist and .whl
- 49. $ hatch release
- 50. So...which tool should I use?
- 52. It depends...
- 53. • building a library? • or an application? • does it fit my infrastructure?
- 54. Thanks