Missing wheels for readthedocs environment

Describe the bug

Since the release of 0.24 readthedocs crashes during the installation of scikit-learn. Prior to 0.24, it would simply install scikit-learn using the available wheel scikit_learn-0.23.2-cp37-cp37m-manylinux1_x86_64.whl. There are no wheels for manylinux1 since 0.24, only manylinux2010. I guess that is why readthedocs fails to find a compatible wheel and try installing from source instead.

Complete logs for last successful build of readthedocs: https://readthedocs.org/projects/orion/builds/12473944/

Installed /home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/orion-0.1.10.post1+g9f8a4aa-py3.7.egg
Processing dependencies for orion==0.1.10.post1+g9f8a4aa
Searching for scikit-learn
Reading https://pypi.org/simple/scikit-learn/
Downloading https://files.pythonhosted.org/packages/f4/cb/64623369f348e9bfb29ff898a57ac7c91ed4921f228e9726546614d63ccb/scikit_learn-0.23.2-cp37-cp37m-manylinux1_x86_64.whl#sha256=da8e7c302003dd765d92a5616678e591f347460ac7b53e53d667be7dfe6d1b10
Best match: scikit-learn 0.23.2
Processing scikit_learn-0.23.2-cp37-cp37m-manylinux1_x86_64.whl
Installing scikit_learn-0.23.2-cp37-cp37m-manylinux1_x86_64.whl to /home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages
writing requirements to /home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/scikit_learn-0.23.2-py3.7-linux-x86_64.egg/EGG-INFO/requires.txt
Adding scikit-learn 0.23.2 to easy-install.pth file

Complete logs of first failing build of readthedocs: https://readthedocs.org/projects/orion/builds/12473944/

Installed /home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/orion-0.1.10.post3+gea90a9e-py3.7.egg
Processing dependencies for orion==0.1.10.post3+gea90a9e
Searching for scikit-learn
Reading https://pypi.org/simple/scikit-learn/
Downloading https://files.pythonhosted.org/packages/f9/98/7a46eb605191918b777796ce5792be83ce9d2a739340be7d21d423fcfaf4/scikit-learn-0.24.0rc1.tar.gz#sha256=d4737b5bd2180c71453b5b1baa47f5ee2ed941d1bdda8eee6a9274d6f9e4b503
Best match: scikit-learn 0.24.0rc1
Processing scikit-learn-0.24.0rc1.tar.gz
Writing /tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/setup.cfg
Running scikit-learn-0.24.0rc1/setup.py -q bdist_egg --dist-dir /tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/egg-dist-tmp-56qidrrm
Partial import of sklearn during the build process.
Traceback (most recent call last):
  File "/tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/setup.py", line 201, in check_package_status
    module = importlib.import_module(package)
  File "/home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1006, in _gcd_import
  File "<frozen importlib._bootstrap>", line 983, in _find_and_load
  File "<frozen importlib._bootstrap>", line 965, in _find_and_load_unlocked
ModuleNotFoundError: No module named 'numpy'
Traceback (most recent call last):
  File "/home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/setuptools/sandbox.py", line 154, in save_modules
    yield saved
  File "/home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/setuptools/sandbox.py", line 195, in setup_context
    yield
  File "/home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/setuptools/sandbox.py", line 250, in run_setup
    _execfile(setup_script, ns)
  File "/home/docs/checkouts/readthedocs.org/user_builds/orion/envs/latest/lib/python3.7/site-packages/setuptools/sandbox.py", line 45, in _execfile
    exec(code, globals, locals)
  File "/tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/setup.py", line 305, in <module>
  File "/tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/setup.py", line 291, in setup_package
  File "/tmp/easy_install-jtb4m3p0/scikit-learn-0.24.0rc1/setup.py", line 227, in check_package_status
ImportError: numpy is not installed.
scikit-learn requires numpy >= 1.13.3.
Installation instructions are available on the scikit-learn website: http://scikit-learn.org/stable/install.html

Any idea why the supported platform wheels changed since 0.24?

Also it looks like I am not the only one affected: https://stackoverflow.com/questions/65776380/readthedocs-trouble-with-sklearn-umap

[astrojuanlu]

Hi @bouthilx and subscribed readers! Just stumbled upon this comment while I was looking for something else.

As mentioned in the docs, it’s either python setup.py install or pip install .. When not using a configuration file (which was your case in master), RTD defaults to the former, which in turn triggered easy_install, I believe. Feel free to open an issue if you find more annoyances or weird things 👍

[bouthilx]

Thanks for the quick response! The build has been consistently failing for more than a month so it was not bad luck. I fixed the issue on my side yesterday by changing the configuration of readthedocs so that it installs with pip: https://github.com/Epistimio/orion/pull/542. That is why it is passing now.

In any case I’m sure adding the manylinux1 wheels will solve the issue for many others, so I would leave this issue open until #19235 is merged.

Thanks again!