Python 2.x not working in virtual environment with Python 3.5 about ipykernel HOT 7 CLOSED

Can you show the contents of /home/user/.local/share/jupyter/kernels/python2/kernel.json?

from ipykernel.

Sure.

{
 "display_name": "Python 2", 
 "language": "python", 
 "argv": [
  "python", 
  "-m", 
  "ipykernel", 
  "-f", 
  "{connection_file}"
 ]
}

from ipykernel.

Looks like this is causing issues here. What would be a solution in this case? I want to use libraries installed in the active conda environment but also keep both Python 2 and 3 available.

Thanks

from ipykernel.

Yes, so python in kernelspec will always refer to the Python of the notebook server's environment. This is why IPython (following Python's own rewrite of the #! line of scripts) hardcodes the full path to sys.executable at install time. By manually changing this, as was recommended in jupyter/notebook#397 is explicitly requesting the behavior you are seeing. If you want it to launch with a specific Python, you must specify the absolute path. If you want it to launch with Python 3 or legacy Python, but be sensitive to the env, then you can use python2 or python3.

from ipykernel.

Thanks. Then what do you recommend for new users? Create an environment for Python 2 and another for Python 3, install ipykernel in both, and point the path in kernel.json to the appropriate location inside the environment?

By the way, using python2 might not work because it uses /usr/bin/python and can't find ipykernel.

from ipykernel.

I wouldn't suggest that new users modify any kernel.json. I would recommend that they run python2 -m ipykernel install and python3 -m ipykernel install one time, and be done. They can do this in envs if they wish, or not. If they choose to create new envs and want to install kernels for those envs, run python -m ipykernel --display-name 'Python <env>' in the env. This can be done on env creating with virtualenvwrapper, if it's a common enough action.

from ipykernel.

Got it. python -m ipykernel --display-name 'Python <env>' looks really useful. Thanks.

from ipykernel.