Some answers:

>I updated my (anaconda) jupyter and something changed. Jupyter Notebook no
longer has a selection of Python 2 or 3 kernels (or other kernels I may
have installed like R or Matlab). It has a selection of all my anaconda
environments.

Yep, that is the nb_conda_kernels extension working.
Can you please confirm that you are not seeing any of the other
pre-existing kernels available in addition to the anaconda environments?

A little bit of explanations: nb_conda_kernels is actually creating
kernelspecs on the fly for each conda environment where you have jupyter
installed.

>I have a notebook that was created with a kernel in one conda environment.
But I want to run it in a different environment. When i start Jupyter
Notebook in that new environment, and then run the notebook, it starts a
kernel from the environment I created the notebook in, rather than the
environment in which I started Jupyter Notebook.
I can see no way to change the kernel in which a notebook runs. Is there a
way?

The main idea behind nb_conda_kernels is being able to start a notebook
server and quickly change between conda environments without killing the
notebook sever, changing the environment and starting the notebook server
again. You JUST select the "environment" from the Kernel menu and you are
done, you are ready to work in that environment in just one step.

>I'm a bit annoyed that Anaconda has made this specific change, because now
you get a different experience depending on whether you install Jupyter
from Anaconda or by other means. Adding some extra features would be fine,
but it sounds like this uses a different model of having multiple installed
kernels.

There was discussion about this and the final decision was to ship a set of
extensions/nbextensions by default in Anaconda...in this way, the Anaconda
users could have a better experience with an Anaconda-Jupyter integrated
experience.

>Maybe we need to do what we've been putting off since we designed
kernelspecs and properly integrate environments, so that Anaconda doesn't
feel the need to stuff environments into the kernelspec mechanism.

That could be interesting... but from the Jupyter side, I am not sure if
Jupyter should do that because it will bring us a lot of complexity derived
from the fact that different "environment" mechanism exists and we will
have to deal with all the upstream problems and diffrences that comes from
those ones...

>Heh. This answers my other question I didn't ask. Is this feature a
customization by Anaconda or is Jupyter smart enough to figure out conda
envs? So yes, I too am not entirely thrilled that I'm not getting the same
experience. Makes me want to reinstall everything from conda-forge.

You don't need to install everything from scratch... just conda remove
nb_conda_kernels and you have the default experience.

Hope this info helps a little bit...

-- 
*Damián*

-- 
You received this message because you are subscribed to the Google Groups 
"Project Jupyter" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jupyter/CAH%2BmRR2jhPdG27KfSkgD4%3DA0pGdwhS%2BHdH-%2Bti0zjgNJueZtWg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to