scikit-learn 0.20.0 breaks sklearn-genetic 0.2

[john-sandall]

Environment Versions

1. OS Type: macOS Mojave (10.14.6)
2. Python version: Python 3.6.8
3. pip version: pip 18.1
4. sklearn-genetic version: 0.2

Steps to replicate

1. I'm using pyenv and virtualenv to control Python environment, so for exact replication, pyenv shell 3.6.8 and mkvirtualenv sklearn-genetic-test to create fresh virtualenv.
2. pip install sklearn-genetic==0.2
3. In a Python shell, run from genetic_selection import GeneticSelectionCV

Expected result

Running from genetic_selection import GeneticSelectionCV works without error.

Actual result

(sklearn-genetic-test) ➜  sklearn-genetic-test python
Python 3.6.8 (default, May 22 2020, 14:02:32)
[GCC 4.2.1 Compatible Apple LLVM 11.0.0 (clang-1100.0.33.17)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from genetic_selection import GeneticSelectionCV
/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/sklearn/utils/deprecation.py:143: FutureWarning: The sklearn.metrics.scorer module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.metrics. Anything that cannot be imported from sklearn.metrics is now part of the private API.
  warnings.warn(message, FutureWarning)
/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/sklearn/utils/deprecation.py:143: FutureWarning: The sklearn.feature_selection.base module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.feature_selection. Anything that cannot be imported from sklearn.feature_selection is now part of the private API.
  warnings.warn(message, FutureWarning)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/genetic_selection/__init__.py", line 32, in <module>
    from sklearn.externals.joblib import cpu_count
ModuleNotFoundError: No module named 'sklearn.externals.joblib'
>>> quit()
(sklearn-genetic-test) ➜  sklearn-genetic-test pip -V
pip 18.1 from /Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/pip (python 3.6)
(sklearn-genetic-test) ➜  sklearn-genetic-test python -c "from genetic_selection import GeneticSelectionCV"
/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/sklearn/utils/deprecation.py:143: FutureWarning: The sklearn.metrics.scorer module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.metrics. Anything that cannot be imported from sklearn.metrics is now part of the private API.
  warnings.warn(message, FutureWarning)
/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/sklearn/utils/deprecation.py:143: FutureWarning: The sklearn.feature_selection.base module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.feature_selection. Anything that cannot be imported from sklearn.feature_selection is now part of the private API.
  warnings.warn(message, FutureWarning)
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/Users/john/.virtualenvs/sklearn-genetic-test/lib/python3.6/site-packages/genetic_selection/__init__.py", line 32, in <module>
    from sklearn.externals.joblib import cpu_count
ModuleNotFoundError: No module named 'sklearn.externals.joblib'

[john-sandall]

This is happening because in scikit-learn 0.20.0 they relocated sklearn.externals.joblib to from sklearn.utils._joblib import cpu_count (see https://scikit-learn.org/dev/whats_new/v0.20.html#miscellaneous).

The easiest fix here is to simply state that sklearn-genetic only works with scikit-learn>=0.20.0. If you want to add backwards compatibility to 0.18 then you'll need some conditional logic to check the sklearn version in order to determine where to import cpu_count from.

For now I can give you a PR which gets this library working again and pip installable via PyPI. There are further issues for getting this working with Python 3.8 which I shall detail in a separate issue.

[john-sandall]

Another reason to simply require scikit-learn>=0.20.0 is because scitki-learn version 0.18 doesn't follow modern Python packaging conventions in that it doesn't specify or automatically install numpy and scipy as required sub-dependencies, instead requiring the user to manually install these in their environment first.

See scikit-learn/scikit-learn#7867 and scikit-learn/scikit-learn#8242 for discussion of this problem and why it existed. This was resolved in sklearn 0.20 - https://github.com/scikit-learn/scikit-learn/blob/0.20.0/setup.py

In certain situations, this may cause issues with even installing sklearn-genetic==0.2, e.g. if someone does indeed want to use scitkit-learn==0.18 and uses pip-tools to pin the version, pip-tools won't put numpy or scipy in it's auto-generated requirements.txt due to sklearn 0.18 not correctly specifying them as requirements, which means the installation fails. This was an issue I faced when trying to test to see how hard it would be to make this package work for both 0.18 and >=0.20.0

[adajinjin]

Similar issue here. It looks like with higher version of sklearn this issue still exists.
Python version: 3.7.7
scikit-learn==0.23.0
sklearn-genetic==0.2
Same error message:
ModuleNotFoundError Traceback (most recent call last)
in
----> 1 from genetic_selection import GeneticSelectionCV

~/opt/anaconda3/envs/recommender/lib/python3.7/site-packages/genetic_selection/init.py in
30 from sklearn.metrics.scorer import check_scoring
31 from sklearn.feature_selection.base import SelectorMixin
---> 32 from sklearn.externals.joblib import cpu_count
33 from deap import algorithms
34 from deap import base

ModuleNotFoundError: No module named 'sklearn.externals.joblib'

[john-sandall]

@adajinjin I need to make a couple changes to #12 after which I think @manuel-calzolari will accept #12 and maybe #13. In the meantime, you can always install from my fork here: https://github.com/john-sandall/sklearn-genetic/

[manuel-calzolari]

Closed with commit c029c0e.