Making Jupyter notebooks less awful

Track: Data Science

Type: Remote Talk

Room: Talk Room 2 (Remote Talks)

Time: Oct 06 (Fri): 09:45

Duration: 0:45

Love them or hate them, Jupyter Notebooks have become an indispensable tool for data exploration, analysis, model development and collaboration. In this talk, we'll try to convince those who love them to use them in a way that those who hate them, hate them less.

We'll start by talking about what makes Jupyter Notebooks indispensable, and what makes them awful. We'll also talk about the various ways people use notebooks. As part of this we'll do a whirlwind tour through IDE options, from local Jupyter in your browser through to the many managed cloud-based options available today.

Then we'll describe some of the patterns and tooling you can use to make your Notebooks less awful (perhaps even beautiful!). Some principles we'll discuss are: good coding practice and style (using nbqa and other code quality tools), version control and reviewing (using Github, Jupyterlab git extensions, nbdime, nbdev, and ReviewNB), testing (using nbval, nbmake, testbook), dependency management (virtual environments, requirements files, Docker), and automating your code quality efforts (Github actions).

The ecosystem of tooling for Jupyter Notebooks is growing, with new packages and features being developed to help smooth over some of Jupyter Notebooks' limitations. This review will give you a birds eye view of some of the current tooling your team can adopt to improve the quality of your Notebooks!

URLs