jupyterhub and conda environments - jupyter

Question

I'm a little bit confused about relations between conda environments and jupyterhub.
As jupyterhub documentation says it can be installed from conda. So it is possible to use some conda envirnment (for example environment "root") and do "conda install jupyterhub" from it.
In the same environment will "live" jupyter. And installation of nb_conda from in this environment gives ability to select kernels and other conda environments in notebooks.
My question is about software like nbextensions and ipywidgets. Where should they be? In the same environment as jupyterhub or in an environment correponds new notebook?

Answer 1

The relationship between conda and jupyter can be a confusing one. Think of conda as your environment, and jupyter as just any other package. A package that you can start a process with and then serve.
To answer your question, they should be installed in your conda environment. Unfortunately it is a little more complicated than that. These extensions will be available to all users. I haven't personally tested a single user having more extensions in a different environment (if it is possible, but will update answer if I do).
If it helps, this is what the docs have to say for the matter:
To install the jupyter_contrib_nbextensions notebook extensions, three steps are required. First, the Python pip package needs to be installed. Then, the notebook extensions themselves need to be copied to the Jupyter data directory. Finally, the installed notebook extensions can be enabled, either by using built-in Jupyter commands, or more conveniently by using the jupyter_nbextensions_configurator server extension, which is installed as a dependency of this repo.
Assuming you installed the extensions via conda:
conda install -c conda-forge jupyter_contrib_nbextensions
then the --sys-prefix was used, which is good. From the docs:
--sys-prefix to install into python's sys.prefix, useful for instance in virtual environments, such as with conda.
So, to add an extension, the process should look like this:
$ sudo su -
$ pip install fileupload
$ jupyter nbextension install --sys-prefix --py fileupload
$ jupyter nbextension enable fileupload --py --sys-prefix
Since the title is asking about conda environments, I'll go into that a little bit too. I've tested these methods on Ubuntu 18.04LTS.
Very often you will want to allow users to share user created environments, never having access to root privileges. You have two good options that I have seen (someone please comment if you know other methods): 1) share an environment 2) duplicate an environment from requirements files. Don't forget you'll have to add the environment as a kernel as well.
method 1 - shared environment
Create a environment in a shared location, and then have both users add it as a kernel.
conda create -p /home/envs/test --clone root
can clone root to copy the root env, or base for the base environment. /home/envs/test will create a "test" environment, in the "envs" directory. Make sure envs has all necessary permissions for users that will be using the files.
From there as another user, just add the environment as a kernel.
$ sudo su - <user-to-install-kernel-to>
$ conda activate <test>
$ python -m ipykernel install --user --name test \
--display-name "Python (test)"
Note.. I believe I had to update the kernelspec manually to get it to point to the correct python environment 🤦‍♂️
method 2
Alternatively, just create a copy of the environment
$ conda env export --name test > environment.yml
$ sudo su - customer
$ conda env create --name cust-env-copy --file environment.yml
$ python -m ipykernel install --user --name cust-env-copy \
--display-name "Python (test)"