Jupyter Notebook/Conda best practices

[Paul Hobson]

Jupyter team:

Congrats on the 5.0 release. The new notebook is sooooo nice. All of the hard work is very much appreciated.

I'm curious what folks (users and devs alike) have settled on as their best practices for managing jupyter and ipykernels when you have many conda environments. 

At the moment I see two general approaches:

1) install everything into every conda environment where they are needed, and launch the notebook server from that environment

or 

2) only install jupyter and notebook into the root (default) conda environment, install ipykernel into the remaining environments, and launch jupyter from the default env, selecting the kernel as necessary.

Is either one of those a particularly bad idea? Is there a different approach I missed?

-Paul

[Matthias Bussonnier]

Hi Paul.

Up until recently, conda was shipping with nb_kernel_extension (or I
don't remember the name) that would automatically discover kernel in
multiple env. So 2 is IMHO (prefered).

Also at some point the notebook _server_ will drop Python 2, so you
will have to run the server from a separate env from your Python 2
kernels. That basically whal all the non-python kernel are doing.

So 2 seem to be the prefered way if the only thing you care about are
kernels extensions.

Nuance on your "2", the notebook does not need to be in the root env,
it can be on any env, and see kernels from other envs.
--
M

> --
> 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 jupyter+u...@googlegroups.com.
> To post to this group, send email to jup...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/jupyter/CADT3MEDT4Qx7sqFo6CnHrwzcTwS1MBSYtnouVT5ku-BqSW9-kw%40mail.gmail.com.
> For more options, visit https://groups.google.com/d/optout.

[Paul Hobson]

Thanks, Matthias. That's helpful.

As a quick followup, I need to do:

python - m ipykernel install --name=<envname_or_similar> 

Within each conda environment that I want available, right?

-paul

> email to jupyter+unsubscribe@googlegroups.com.

> To post to this group, send email to jup...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/jupyter/CADT3MEDT4Qx7sqFo6CnHrwzcTwS1MBSYtnouVT5ku-BqSW9-kw%40mail.gmail.com.
> For more options, visit https://groups.google.com/d/optout.

--
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 jupyter+unsubscribe@googlegroups.com.

To post to this group, send email to jup...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/jupyter/CANJQusVHC05Y5nZmNdfjDh0HAy71C%3D6kYtg%2B69AgOZ7rJpVLGg%40mail.gmail.com.

[Thomas Kluyver]

On 7 April 2017 at 23:16, Paul Hobson <pmho...@gmail.com> wrote:

As a quick followup, I need to do:

python - m ipykernel install --name=<envname_or_similar> 

Within each conda environment that I want available, right?

Yup - that or use Continuum's plugin for the notebook to discover conda environments. That comes with some of its own headaches, though.

I'm trying to come up with a better mechanism for Jupyter to know about multiple kernels for the same language.
https://github.com/jupyter/nbformat/issues/81

[Jan Schulz]

Hi,

On 6 April 2017 at 23:53, Matthias Bussonnier

<bussonnie...@gmail.com> wrote:
> Up until recently, conda was shipping with nb_kernel_extension (or I
> don't remember the name) that would automatically discover kernel in
> multiple env. So 2 is IMHO (prefered).

https://github.com/Cadair/jupyter_environment_kernels can do that as
well: install it (from git master), enable it in the notebook config
and you get support for conda (python and R kernels) and virtualenv
environments, including activating the environment before starting the
kernel.

Kind regards,

Jan