search

schollz / find3

High-precision indoor positioning framework, version 3.
https://www.internalpositioning.com/doc
MIT License
4.63k stars 364 forks source link

Make requirement less strict for numpy #168

Closed Wykks closed 5 years ago

Wykks commented 5 years ago

Allow usage of system installation of numpy.

I made a updated Dockerfile without mosquitto (I will probably share it if interested) for raspberry (using external mosquitto). Because the requirements are too strict, pip doesn't pick up the version installed by apt (v1.16.2 in my case, from https://packages.debian.org/buster/python3-numpy).

Numpy is hard to compile on a raspberry, and even failed in my case for some reasons.

schollz commented 5 years ago

Thanks looks good!

Wykks commented 5 years ago

@schollz hi! Sorry for the delay here, I marked this pr as draft because I wasn't sure about allowing <2.0.

I can confirm it's working with v1.16.2, but just FYI, I'm getting this warning:

usr/local/lib/python3.7/dist-packages/sklearn/ensemble/weight_boosting.py:29: DeprecationWarning: numpy.core.umath_tests is an internal NumPy module and should not be imported. It will be removed in a future NumPy release.
from numpy.core.umath_tests import inner1d

After reading through numpy changelogs, it seems they are making breaking changes even on minor version... So maybe just to be safe, we could just allow 1.16.x, eg: numpy>=1.15.1,<1.17 This at least, allow use of the numpy version packaged in debian buster.