Pin requirements.txt

[tunnell]

If you pin versions, it will make things in the long run easier for reproducibility. Also, we make sure that we are testing the same versions that we are deploying.

[coveralls]

Pull Request Test Coverage Report for Build 13

• 0 of 0 (NaN%) changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 52.248%

---

Totals
Change from base Build 12:	0.0%
Covered Lines:	430
Relevant Lines:	823

---

💛 - Coveralls

[JelleAalbers]

Thanks! While I was against this for pax in the past, it makes sense. We've had a few issues when new numpy and numba versions came at one point. So let's pin.

However we should probably stop filling install_requires from requirements.txt after pinning; see e.g.

• https://packaging.python.org/discussions/install-requires-vs-requirements/
• https://caremad.io/posts/2013/07/setup-vs-requirement/
• Support for setup.py pyupio/pyup#137

The best practice seems to be to have pinned dependencies in requirements.txt and unpinned ones in install_requires. This way you can have consistent deployments with pip install -r requirements.txt while pip install remains flexible, e.g. for people who want to use part of strax in their projects.

Moreover, this allows you to pin dependencies of your dependencies in requirements.txt; essentially requirements.txt can then become just the pip freeze output.

[JelleAalbers]

Maybe we should also look at pipenv, it's apparently now the officially recommended python packaging tool.

[tunnell]

@JelleAalbers I'll look at this install requires thing...

[JelleAalbers]

Cool, I just did JelleAalbers@718456d now, but the repetition feels a bit wrong.