failed to install Scipy via `pip` (OSX M1)

System Environment:

macOS Big Sur 11.5.2, M1 processor Python 3.8.11 (using a venv) pip==21.2.4 (updated via pip) setuptools==57.4.0 (updated via pip)

Observed output

(venv) % python -m pip install scipy
Collecting scipy
  Using cached scipy-1.7.1.tar.gz (36.1 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
    Preparing wheel metadata ... error
    ERROR: Command errored out with exit status 1:
     command: /Users/work/Repos/endaq-python-calc/venv/bin/python /Users/work/Repos/endaq-python-calc/venv/lib/python3.8/site-packages/pip/_vendor/pep517/in_process/_in_process.py prepare_metadata_for_build_wheel /var/folders/1s/lrjbk2v17m7063f944nmy7vh0000gn/T/tmp7dun4f3u
         cwd: /private/var/folders/1s/lrjbk2v17m7063f944nmy7vh0000gn/T/pip-install-cc3mdfet/scipy_90eec6ff22e54b9d84809dedc230c072
    Complete output (4 lines):
    setup.py:490: UserWarning: Unrecognized setuptools command ('dist_info --egg-base /private/var/folders/1s/lrjbk2v17m7063f944nmy7vh0000gn/T/pip-modern-metadata-0oa55ptb'), proceeding with generating Cython sources and expanding templates
      warnings.warn("Unrecognized setuptools command ('{}'), proceeding with "
    Error: 'None' must be installed before running the build.

    ----------------------------------------
WARNING: Discarding https://files.pythonhosted.org/packages/47/33/a24aec22b7be7fdb10ec117a95e1e4099890d8bbc6646902f443fc7719d1/scipy-1.7.1.tar.gz#sha256=6b47d5fa7ea651054362561a28b1ccc8da9368a39514c1bbf6c0977a1c376764 (from https://pypi.org/simple/scipy/) (requires-python:>=3.7,<3.10). Command errored out with exit status 1: /Users/work/Repos/endaq-python-calc/venv/bin/python /Users/work/Repos/endaq-python-calc/venv/lib/python3.8/site-packages/pip/_vendor/pep517/in_process/_in_process.py prepare_metadata_for_build_wheel /var/folders/1s/lrjbk2v17m7063f944nmy7vh0000gn/T/tmp7dun4f3u Check the logs for full command output.

It then tries other versions of scipy, but they all result in the same message.

Any help appreciated, thanks!

[CrepeGoat]

More fixes for anyone else with the same problem (and as a reference for myself, if we’re being honest):

Good to know, thanks! For anyone else reading, I’ll probably be pip installing directly from the git master branch via pip install git+https://github.com/scipy/scipy@master

So this resulted in some errors for me due to, afaik, missing c-library dependencies when compiling from source.

Now I could dig into these C compilation issues. But I’m a Python developer on a time budget, so I’m going to instead opt for the easier solution.

The easier solution

Check your Python interpreter

If you install python through the installer available directly from https://www.python.org/, you’ll install a universal binary. You can tell it’s a universal binary because running in the command line which python3 | xargs file prints to stdout something like the below message, which makes note of both x86_64 and arm64 architectures:

crepegoat@CrepeScape ~ % which python3 | xargs file
/usr/local/bin/python3: Mach-O universal binary with 2 architectures: [x86_64:Mach-O 64-bit executable x86_64] [arm64:Mach-O 64-bit executable arm64]
/usr/local/bin/python3 (for architecture x86_64):	Mach-O 64-bit executable x86_64
/usr/local/bin/python3 (for architecture arm64):	Mach-O 64-bit executable arm64

(If you installed via another means like e.g. Homebrew, as long as the above command prints the same result this will still work.)

Set zsh to run programs in x86_64 mode

By default, Terminal (i.e., zsh) will run programs in arm64 mode; this is evidenced by the result of running uname -m:

crepegoat@CrepeScape ~ % uname -m
arm64

This can be changed on either a command-, terminal-session- or permanent-basis using the arch command. Below I opt for the session-basis change:

crepegoat@CrepeScape ~ % arch -x86_64 zsh
crepegoat@CrepeScape ~ % uname -m                          
x86_64

Install scipy as usual

Now after setting up zsh in x86_64 mode, running pip will install binaries for the x86_64 architecture, with which scipy has perfectly functional compatibility. At the time of writing this, I’ve successfully installed scipy==1.7.1 on my system with this method. If you’re reading this, I hope it helps you too 😃

[GT135813]

@CrepeGoat No worries at all. Best of luck.