Closed jmsdnns closed 6 years ago
Sounds great but does it still work with conda without defining site_packages?
I just had success doing it.
My specific steps were:
conda create --name beakerx python=3.6
source activate beakerx
cd path/to/this/repo
pip install -e .
beakerx-install
jupyter nbextension enable beakerx --py --sys-prefix
jupyter nbextension enable --py widgetsnbextension
I turned it on and ran code in a Python 3 notebook with success.
I thought I'd grep ~/.conda
for getsitepackages
to see what'd turn up and it seems conda is also packaging this file with their distributions. Their version implements getsitepackages
. That explains why it only came up with virtualenv
.
I know
conda
is the officially supported way to install beakerx. This is consistent with norms inside the data science world. Everyone else in the Python community, however, uses virtualenv's, as it is the official environment system from Python.Virtualenv supplies its own
site.py
and does not implementgetsitepackages
. This causes an issue becausebeakerx/beakerx/setupbase.py
attempts to call it.This line is in
setupbase.py
.The
site_packages
variable is never used, so this line serves no purpose, as far as I can tell, and the variable it creates is never referenced.Removing the line is all that's required for virtualenvs to work.