Biopython and Scikit-learn (requirements.txt) both require numpy and scipy respectively but do not explicitly list these as install_dependencies in their setup.py files. Both packages argue that numpy and scipy should be installed abstractly as part of a scientific server.
To comply I think it would be best installing these two libraries as part of a raw server provisioning using apt-get:
There are a number of requirements that are essential for this package and, I believe, belong in the setup.py install_requires clause:
install_requires=['numpy>=1.11.0', 'pandas>=0.17.1', 'futures>=3.0.5', 'decorator>=4.0.9', 'biopython>=1.66', 'taxtastic>=0.5.7', 'scipy>=0.17.0']
Biopython and Scikit-learn (requirements.txt) both require numpy and scipy respectively but do not explicitly list these as install_dependencies in their setup.py files. Both packages argue that numpy and scipy should be installed abstractly as part of a scientific server.
To comply I think it would be best installing these two libraries as part of a raw server provisioning using apt-get:
apt-get install python-numpy python-scipy(https://www.scipy.org/install.html)or pip:
pip install numpy scipyWe can then correctly state our Deenurp install_requires and requirements.txt files for abstract and hard dependencies.