JupyterCon 2023

WAAAT! Accessibility Testing JupyterLab
05-12, 15:30–16:00 (Europe/Paris), Louis Armand 2

WAAAT = Web App Automated Accessibility Testing


Have you ever wondered how to make a complex web application accessible to disabled users? There are many pieces to this puzzle, which include things like using semantic HTML, consulting disabled users, and testing your app before releasing it. In this talk, we will look at testing, specifically how you can use existing tech to increase the portion of testing that can be automated, while keeping in mind that you will always at some point need to test with disabled users.

When it comes to accessibility testing, one of the challenges—and opportunities—in working with JupyterLab is that well known existing solutions are geared towards low-interaction web pages rather than highly interactive web apps. The better part of my work for over a year has been working with designers and developers in open source to develop an accessibility testing solution capable of dealing with the complexity of JupyterLab.

In this talk, I am going to do a deep dive of this testing solution as well as the challenges faced and lessons learned. I will show-and-tell a testing system combined with a documentation system, inspired in part by the work of the W3C Accessibility Conformance Testing (ACT) group. Most importantly, I will demonstrate how two tools, Playwright and GitHub Actions, can be combined to create a powerful CI/CD system for automated accessibility testing. It's not magic. You still have to write the tests yourself. But it's a powerful combination that you can learn by coming to this talk. You can then adopt and adapt this system to many other complex web apps when you need to get serious about making them accessible to disabled users.

Intended audience

Anyone curious about automated UI testing and accessibility! Some familiarity with web accessibility will be useful—for example, knowing what WCAG and Pa11y are; if an attendee doesn't already know these terms, attending any other accessibility talk before this one should give all of the accessibility background needed for this talk. Knowledge of Playwright and GitHub Actions is not necessary, but a little bit of JavaScript web programming experience, as well as a nominal understanding of CI/CD, will make following some parts of the talk easier.

Talk outline

  1. Introduction (5 minutes)
    - Introduction to accessibility testing in the context of JupyterLab
    - Tools of the trade for accessibility testing
    - Intro to testing and reporting: automated versus manual, auditing versus regression prevention, etc.
  2. Challenges faced (5 minutes)
    - The limitations of axe-core, pa11y and the like for JupyterLab
    - Lessons learned so you do not have to
  3. A proposed framework for automated accessibility testing (5 minutes)
    - Use axe-core, but also use Playwright to write more powerful and specific tests
    - Focus on regression testing
    - Mapping tests to WCAG guidelines
  4. Call to action (5 minutes)
    - How to make fixes and keep them fixed!
    - How to use the system to accompany an accessibility fix with a regression test
  5. Questions and answers (10 minutes)


At the end of this talk, attendees will:

  • Know how to write accessibility regression tests for JupyterLab, where to add them, and how to get help if needed.
  • Have an example of an automated accessibility testing system that they can pattern off of to bring an accessibility testing practice to other web apps.
  • Have a better understanding of the current state of the art for automated UI testing in the browser and how to apply it to automated accessibility testing.

Gabriel Fouasnon is a frontend caterpillar (10 years experience) morphing into a web accessibility butterfly (1 year in the making). He works at Quansight Labs under a grant from the Chan Zuckerberg Initiative’s Essential Open Source Software for Science program to improve accessibility in the Jupyter ecosystem. (Follow along at jupyter-a11y.netlify.app!) This is his first conference talk.