@deathbeds/jyve

Jupyter Kernels in the Browser


Keywords
jupyter, jupyterlab, jupyterlab-extension, jyve
License
BSD-3-Clause
Install
npm install @deathbeds/jyve@0.7.0

Documentation

jyve

Build Status Binder npm-version

Experimental, unsafe, interactive Jupyter Kernel-like things in your browser. Try the demo.

IFrame-backed kernels

Convenience wrappers

  • [PhosphorJS](./notebooks/Phosphor Playground.ipynb)
  • [d3](./notebooks/d3 Playground.ipynb)

These kernels create their own widget next to a Notebook (or Console). Restarting the kernel is equivalent to refreshing the page.

The Big Security Hole

For extra danger, these kernels also make the root JupyterLab application instance available. In particular, this allows you to do things like:

JupyterLab.commands.execute('notebook:create-new');

...to create a new notebook, though you can do [just about anything](./notebooks/JupyterLab API.ipynb).

Before

Install:

  • JupyterLab >=0.31.10 from pip or conda

Install

# the core manager, required but doesn't do anything by itself
jupyter labextension install @deathbeds/jyve
# the base kernel
jupyter labextension install @deathbeds/jyve-kyrnel-js-unsafe
# specific compile-to-js kernels (needs the js kernel)
jupyter labextension install @deathbeds/jyve-kyrnel-brython-unsafe
jupyter labextension install @deathbeds/jyve-kyrnel-coffee-unsafe
jupyter labextension install @deathbeds/jyve-kyrnel-p5-unsafe
jupyter labextension install @deathbeds/jyve-kyrnel-typescript-unsafe
# extra packages, wrapped for convenience in jyve kernels
jupyter labextension install @deathbeds/jyve-lyb-d3
jupyter labextension install @deathbeds/jyve-lyb-phosphor

Or, since hey, This is Jyve:

jupyter labextension install \
    @deathbeds/jyve \
    @deathbeds/jyve-kyrnel-brython-unsafe \
    @deathbeds/jyve-kyrnel-coffee-unsafe \
    @deathbeds/jyve-kyrnel-js-unsafe \
    @deathbeds/jyve-kyrnel-p5-unsafe \
    @deathbeds/jyve-kyrnel-typescript-unsafe \
    @deathbeds/jyve-lyb-d3 \
    @deathbeds/jyve-lyb-phosphor \
  && jupyter labextension list

Motivation

JupyterLab currently disables or limits the scope of of arbitrary JavaScript in output cells, Markdown documents and other places. This is a Good Thing, and will help keep safe people who are primarily interested in learning and doing science, research, and analysis and don't want to be bothered with cross-site scripting, click-jacking, privilege escalation and the other things that a Jupyter kernel and attached JavaScript Jyve session can do.

However, one of the beautiful elements of the Classic Notebook was the ability to "build the ship while you are sailing it," e.g. writing the Python and JavaScript for a new ipywidget in the same notebook, writing notebook extensions. The IPython %%javascript magic, for example, will just emit a nasty warning in JupyterLab. However, this immediately exposed you to a relatively high likelihood of breaking other extensions, or even core behavior itself.

Jyve fits somewhere between the two.

Kyrnels

A Jyve Kyrnel runs in JupyterLab and has full, unsafe access to the many of the capabilities of your browser, including:

  • its own dedicated DOM in an iframe
  • the JupyterLab Application instance, including
    • commands
    • the application shell
    • so much more...
  • debugging with in-browser (or remote) tools
  • localStorage, IndexedDB and other storage mechanisms

Because it's almost a real Jupyter Kernel, a Jyve Kyrnel can be used by tools like the JupyterLab Notebook and the JupyterLab Console. But, because of its relationship to JupyterLab and the browser, it can:

  • load arbitrary code and data from anywhere on the internet
  • integrate with the excellent local browser debugger tools
  • run JupyterLab commands
  • add new phosphor Widgets to the application shell

Lybs

JupyterLab, while very customizable, is still delivered as a series of intentionally-unpredictable files which bundle many megabytes of JS, CSS, JSON and other artifacts.

To make it easier to learn about key libraries inside JupyterLab (or even external to it), a Jyve Lyb wraps up a library and makes it usable directly inside every kyrnel.

Development

Before

Install:

Setup

conda env update
source activate jyve-dev
./postBuild

Build Once

jlpm build
jupyter lab build

Always Be Building

jlpm watch
# and in another terminal
jupyter lab --watch