Modern development environments for Pythonistas

Speaker

Dustin Ingram

Material

• video

Note

• Speaker

  • Dustin Ingram (twitter)
    • Organize PyTexas
    • PyPi contributor

• Topics

  • dev, test env
  • dependency workflows
  • linting & autoformatting
  • go to prod

• Focus: anything that's not code and will be check into code base

dev, test environment

Your development environment.
"But it works on my machine"

You laptop is not production.

• Problem: virtualenv only isolate Python. Platform-level stuff are still shared.

• Fully isolate a reproducible environments

  • containerization (e.g., docker)
    • Dockerfile
    • docker-compose

A modern workflow for dependencies

• Problems: When install a new package and it depends on newer version of old dependency, it might upgrade unnoticeably.

• Don't pip freeze > requirements.txt anymore

  • Bad
    • Get everything from your system
  • Good
    • Version pinning
    • Managing sub-dependencies

pip-tools

• pip-tools

• Pin and compile the dependencies and sub-dependencies
  sh pip-compile requirements.in

• Pinning and compiling only get 90%
  • Artifact hashing
    • pip compile --generate-hashes requirements.in
    • pip install --require-hashes -r requirements.txt
• Now our deps are 100% frozen

When / How to upgrade dependencies

• When: early and often
• How: automated dependency services → scan you repo and see your deps
  • pyup.io
  • dependabot

Linting and autoformatting

Use black

Taking your env to prod

Through services that supports deployment through container

Conclusion

1. Use docker
2. Use docker-compose
3. Use pip-compile from pip-tools
4. Use dependabot / pyup.io
   for auto-upgrade deps
5. Use dependabot
6. Deploy your docker container to prod

Q & A

• Q: pipenv, poetry or pip-tools
  • A: Personal preference

Share on: Twitter ❄ Facebook ❄ Email

---