Description
Hi there! I've moved away from nb_conda_kernels due to... so many conflicting dependencies and CI/CD nightmares. But the pip install coconut[kernel] technique still works great for this workflow, which makes use of pyenv/virtualenv. Jupyter/lab can find the kernel just fine, provided a small kernel.json config file. I figured out how to add it in a way that duplicates those workflows.
Proposal
Request: add a coconut kernel install [--name] [--user] cli command
functionality: creates a ~/.local/share/jupyter/kernels/{myenv}-coconut/kernel.json spec file, with the contents:
{
"argv": [
"${PYENV_ROOT}/versions/{myenv}/bin/python",
"-m",
"coconut.icoconut",
"-f",
"{connection_file}"
],
"display_name": "{myenv}-coconut",
"language": "coconut"
}Where the {PYENV_ROOT} and {myenv} variables get replaced with the pyenv root dir and currently-activate environment name, respectively.
For now I just copy+edit that file around to new kernel folders every time I make one that uses Coconut, but that would be easier if there was a coconut equivalent to the python -m ipykernel install --user --name {myenv} command that I run for python kernels that makes the kernel.json automatically.
Assumptions: The user wants a coconut kernel to have the same name as the python environment it's originating from, but with an added -coconut in the name. Seemed good enough?