Same problem as in this question sys.path different in Jupyter and Python - how to import own modules in Jupyter?. In pure Python, it prepends my system environment variable PYTHONPATH to sys.path but Jupyter notebook doesn't, so I can't import my own module.
There are many similar questions asked on SO, and the solution is to directly manipulate sys.path in the script.
Is there a way to make Jupyter notebook use my system PYTHONPATH variable, as in pure python?
Use simply the PYTHONPATH.
export PYTHONPATH=/Users/user/my-other-library/
jupyter notebook
I just tested with the newest jupyterlab-2.1.2 and it works.
JupyterLab reuses the PYTHONPATH on Linux, so I created a file like
#!/bin/bash
# add your path
export PYTHONPATH="$PYTHONPATH:/opt/your/path"
# start JupyterLab using an environment
/opt/anaconda/envs/MY_ENVIRONMENT/bin/jupyter-lab
saved it as start_my_jupyterlab, make it executeable with chmod a+x start_my_jupyterlab and run it on the shell with start_my_jupyterlab.
/Users/ME/anaconda3/envs/MY_ENVIRONMENT/bin/jupyter-labJupyter uses its own JUPYTER_PATH environment variable.
--Just chiming in here since the accepted answer didn't give the complete solution--
You can add the path to your modules to the JUPYTER_PATH environment variable, just the same as you would for modifying the PYTHONPATH environment variable:
export JUPYTER_PATH="${JUPYTER_PATH}:/path/to/add/here/"
If you're on a Mac or other Unix system, you would just drop the above line into your ~/.bash_profile
Hint: make sure you run source ~/.bash_profile to enact the changes and close and restart your jupyter notebook.
You can just add the environment variable to the kernel.json file of the used kernel (virtual env):
,"env": {
"PYTHONPATH": "/path/to/folder:$PYTHONPATH"
}
The kernel.json could be found somewhere in the .local folder (~/.local/share/jupyter/kernels/your_venv/kernel.json)
