Discuss a selection of "default" Jupyter extensions in Anaconda

Anaconda (http://continuum.io/download) is a major avenue for people to get started with Jupyter. We, at Continuum, believe Jupyter lowers the bar significantly for a wide range of people who otherwise wouldn’t have access to the tools and interfaces Jupyter presents. We observe two important things about this “new user community” (in contrast to “power users”):

• They get updates of Jupyter by downloading our quarterly major release of Anaconda
• They never want to do anything at the command line, and certainly no operations that they perceive may break their Anaconda (and by implication: Jupyter) software. This means:
  • They won’t do conda install or pip install or git anything
  • They won’t do conda create or virtualenv venv
  • They will be very hesitant doing any bang operations inside of Jupyter

Our response to this situation is that we want to provide even more “out of the box” goodness by default (extending the idea of the rich set of Conda packages we already provide – over 200), and we’d like to include ~half-dozen notebook extensions pre-installed, pre-configured, and pre-enabled with the next release of Anaconda in mid/late Spring 2016. We expect this will ship with Jupyter 4.2 (or possibly 4.3 if it is complete by the change-freeze date for the Anaconda Release – typically 2 weeks prior to target release).

We want to make sure everyone understands that “virgin” Jupyter would always be available to Anaconda users by at least these two methods:

• miniconda + conda install jupyter
• conda create -n vanilla_jupyter jupyter

Which is to say, the Jupyter package we create will continue to be “just vanilla Jupyter + its core dependencies”

We would like to get input on:

• Is the Jupyter community generally supportive of this idea?
• Are there any considerations we need to think about over the next 4-8 weeks leading up to this?
• Which extensions should we include.

On the last point, we’d like to start with a fairly short and conservative set (perhaps a half dozen extensions), and focus on selecting extensions from these two sets:

• Extensions already tracked here: https://github.com/ipython-contrib/IPython-notebook-extensions#documentation
• Extensions that provide a more cohesive user experience for the Anaconda user, providing Jupyter-internal GUI elements for conda-related operations, access to Anaconda Cloud, and creation of rich presentations.

Continuum’s selection criteria right now (but these are open for discussion!) are to bias towards the small set of extensions that are:

• generally applicable (most Jupyter users will find them useful)
• robust (well tested, widely used)
• cross-platform (Windows, Mac, Linux)
• minimal or no dependencies beyond what Anaconda already provides or can be quickly, easily, and reliably fetched or shrink-wrapped (NPM packages).
• utilize the 4.2 extension API for uniform extension deployment, configuration, and enable/disable.

[bollwyvl]

@jcb91 https://github.com/jcb91/jupyter_nbextensions_configurator looks great! I’ll have to look into it more!

[nicolaskruchten]

@bollwyvl ok, I finally have some time to devote to PivotTable.js/Jupyter integration work… Are there any resources anywhere I could read in order to find out how to “have all of those CDNJS links replaced with a node build stack, and then have those assets installed via normal nbextension mechanisms, etc” ? I’m having troubling finding a starting point… Perhaps a similar extension that I could base my work on?

I’m happy to correspond by email as well!