The Future of the Jupyter Notebook

PyConDE & PyData Berlin 2023

2023-04-19

Jeremy Tuloup

• https://github.com/jtpio
• https://twitter.com/jtpio
• https://fosstodon.org/@jtp

About

Technical Director at QuantStack

Core Jupyter Developer

History

Interactive Computing

• Compose
• Submit
• Run
• Output
• View
• Repeat

IPython: the Terminal as a UI

• The terminal is a user interface!
• Challenges:
  • No narrative
  • No memory
  • No reproducibility and communication

REPL: Read Eval Print Loop

Why is the Jupyter Notebook useful?

• Maintain full interactive computing workflow
• Adds:
  • narrative
  • memory
  • repoducibility
  • communication

IPython

Jupyter Notebook

• Millions of users
• Millions of notebooks on GitHub
• Heavily used in data, science, education, finance...

Working with a Notebook

• Client / Server Architecture
• Standard file format
• Kernels for execution

The Jupyter Protocol

• Messaging in Jupyter: https://jupyter-client.readthedocs.io/en/stable/messaging.html
• Frontends and kernels speak the same protocol

Classic Notebook

• Released in 2015 as a successor to the IPython Notebook
• Built with jQuery, RequireJS, Bootstrap

JupyterLab

• Next Gen Notebook Interface (and more)
• Extensible Computing Environment
• Built on modern web technologies

The state of the Classic Notebook

• Very few maintainers, very little time
• But new issues were still coming up on the jupyter/notebook repository, but rarely got any replies
• New weekly meeting to triage issues
• [DISCUSS] How to improve supportability of the repo? https://github.com/jupyter/notebook/issues/5360

• https://github.com/jupyter/notebook/issues/6210

The Reality

• Simplicity is a feature (less is more)
• Still lots of users of the Classic Notebook, especially in education
• Document-centric interface helps remove the noise and focus on the content
• The "tabs-in-tabs" can be confusing (the web browser already has tabs)

Problems

• The Classic Notebook has been unmaintained for years
• It was very difficult to extend and make different extensions co-exist
• Built on "old" web technologies (jQuery)
• And still a lot of people using it!

The initial plan was to sunset Notebook

The current plan is to sunrise Notebook

Jupyter Enhancement Proposal 79

• "Build Jupyter Notebook v7 off of JupyterLab components"
• Merged on Dec 24, 2021
• https://jupyter.org/enhancement-proposals/79-notebook-v7/notebook-v7.html

RetroLab

• Started end of 2020
• https://github.com/jupyterlab/retrolab
• Built with JupyterLab components and extensions
• But keep the Classic Notebook look and feel
• "RetroLab running on Jupyter Server will be the basis for Notebook v7."

Keep the Classic Notebook look and feel

New Features

• Many coming from JupyterLab automatically
• Mime extensions
• Debugger
• Accessibility (CodeMirror 6)
• Theming and Dark Mode
• Built-in Table of Contents
• Can be used in JupyterLite
• Real Time Colaboration (demo)

Demo

• Locally: http://localhost:8888/tree
• Try it on Binder:

Extensions

• What about popular extensions like RISE and nbgrader?

JupyterLab Extensions

• Most JupyterLab extensions are compatible with Notebook 7
• Popular Classic Notebook extensions have been ported to JupyterLab 4 / Notebook 7
• https://jupyterlab-contrib.github.io/extensions.html
• https://jupyterlab-contrib.github.io/migrate_from_classical.html

Migration

• JupyterLab 4, Notebook 7 and NBClassic 1 can be used together on the same Jupyter Server (demo)
• Migration guide available in the documentation:
  • https://jupyter-notebook.readthedocs.io/en/latest/migrate_to_notebook7.html
  • Describes several scenarios for running both the Classic Notebook and Notebook 7 on the same server

Migration Banner

• NBclassic and Notebook 6.5.x show a banner to encourage users to migrate to Notebook 7

Server logs

Current State

• Jupyter Notebook 7.0.0b0 is available
• Install it with pip: pip install --pre notebook
• Install it with conda: conda install -c conda-forge/label/notebook_beta notebook
• Please test and report issues and feedback!

Where is development happening?

• Notebook 7 is being developed in https://github.com/jupyter/notebook (main branch)
• The Classic Notebook UI has been moved to https://github.com/jupyter/nbclassic
• Notebook 6.5.x (security patches only) in https://github.com/jupyter/notebook/tree/6.5.x

What's next

• Blog post about Notebook 7 and the Migration Guides
• Notebook 7 Release Candidates
• JupyterLab 4 final release expected for JupyterCon 2023 early May 2023
• Notebook 7 final release expected after JupyterLab 4 (date TBC)

How can I help?

• If you are a user of the Classic Notebook and have concerns, please reach out!
• Please test the JupyterLab 4 and Notebook 7 betas and give feedback
• Report issues and feedback on the Notebook repository: https://github.com/jupyter/notebook/issues
• Start discussions on the Jupyter Discourse: https://discourse.jupyter.org

Thanks!

Special thanks to the Notebook Team and the authors of the JEP:

Eric Charles, Sylvain Corlay, Afshin Darian, Sharan Foga, Eric Gentry, Kevin Goldsmith, Brian Granger, Jason Grout, Fernando Pérez, Zach Sailer, Rosio Reyes, Jeremy Tuloup

And to the broader Jupyter Community!

JupyterCon 2023

May 10-12, 2023 - Cité des Sciences, Paris

jupytercon.com

Resources

You can find the slides on GitHub:

• Repository: https://github.com/jtpio/pyconde-pydata-berlin-2023-notebook-7
• Live version: https://jtpio.github.io/pyconde-pydata-berlin-2023-notebook-7/