Description
Using nb_conda_kernels as my primary workflow (i.e. not needing a separate jupyter notebook server running for every kernel, just installing a kernel in each environment with one central environment running the server) I would
- install jupyter, nb_conda_kernels, etc in the main/base/default environment, and run jupyter.
- install ipykernel, project dependencies, etc in a project-specific environment.
- open jupyter interface, and switch kernel. Not to "python 3", etc, but to
Python [conda env:projectenvname]which gives me all of my packages/dependencies needed for that project, still on the one jupyter server.
My question revolves around using coconut within this workflow (which is becoming more standard in teaching/academic circles afaik). I have coconut[all] installed in both the main and the "cocotest" project environment, and have run coconut --jupyter successfully in both as well. But I still cannot access coconut syntax directly (i.e. no magic functions) within a notebook using the Python [conda env:cocotest] kernel. Given that nb_conda_kernels explicitly supports other languages (ipykernel, r-irkernel, anything on that list of languages above) as long as the kernel is installed inside that environment, i don't really understand why coconut isn't working?
I did also try running coconut --jupyter notebook from the main environment, and selecting the Python [conda env:cocotest] kernel, but still no dice.
I think one solution is to either make a separate installable or maybe optional coconut[kernel] installation that will act like ipykernel, so that users can add coconut support in jupyter to their versioned dependency list (e.g. environment.yml, etc), just like they can now with ipykernel and r-irkernel, while only needing to run a single jupyter server at any given time.