Track Jupyter Lab and Notebook Accessibility

[balajialg]

Summary

Creating this issue to formulate our goals & actionable next steps for what accessibility looks like in Datahub and potential upstream improvements that need to be made to ensure that the service hits the defined campus accessibility milestones. One of the key stakeholders interested in Datahub hitting this milestone is RTL leadership, who have been vocal about us putting extra effort to improve tool's accessibility.

However, as we all know, we don't control the end-user experience of any of the Jupyter tools - Classic Notebook, Jupyter Lab, and Retro Lab. Considering this reality, we must actively evaluate the campus requirements, track whether those improvements get logged in relevant GitHub repositories, and track their development journey. If issues did not get tracked upstream, raise relevant GitHub issues in the respective repository so that the key stakeholders in the Jupyter ecosystem know this need. The main repository that tracks all the accessibility-related efforts in Jupyter is the Accessibility Repo.

Individual repositories tracking accessibility-related stuff include the following repositories,

• Classic Notebook

• JupyterLab

• Jupyterhub

How can we integrate with the upstream work happening in the Jupyter ecosystem to achieve the accessibility goals? Our current strategies are,

• Be an early tester for Jupyter Lab 4.0 pre-release version. Deploy this version as part of a new hub and test for CM 6.0 changes pushed by QuantStack.
• Share our goals with the timeline for the Quantsight folks. Support them interim in order to achieve the set goals.

Useful Documentation

• Quantsight roadmap for adding accessibility-related updates - https://github.com/orgs/Quansight-Labs/projects/5/views/5

• Quantstack PR for upgrading to Code Mirror 6.0 - Migrate to Codemirror 6 jupyterlab/jupyterlab#11638

• Datahub Accessibility Audit - https://docs.google.com/spreadsheets/d/1kpOUkYmy3p4rKiQW72cgrK1HZ_PlVsD2Px60lMGsbms/edit#gid=492520477

• Datahub Accessibility short-term plan - https://docs.google.com/document/d/1_E9Q7bhx2vMwO0LnQS5YLJCszXYrNsgm4HkjooX_-Lw/edit?usp=sharing

Tasks to complete

• Set up a new hub to test Jupyter Lab 4.0 pre-release version which contains the latest update to the Code Mirror editor @yuvipanda
• Share Berkeley's accessibility goals + timeline with Quantsight folks
• Test Jupyter Lab 4.0 when it releases in a11y hub
• Test Jupyter Notebook 7.0 when it releases in a11y hub
• Support Data8x edx course relaunch

[balajialg]

@ericvd-ucb @yuvipanda Notes from 12/16 Accessibility Workshop! As @ericvd-ucb suggested, The next step would be to summarize the learnings about the accessibility improvements happening across the Jupyter ecosystem into a memo that could get shared with the leadership!

[ryanlovett]

@yuvipanda suggested in previous conversation that before the web accessibility clinic reviews datahub again, we should wait for jupyterlab to integrate the new CodeMirror editor (version 6 I presume which "was designed from the start with accessibility in mind"). Otherwise we are likely to see similar results as before.

[ryanlovett]

Here is the jupyterlab issue for integrating CodeMirror 6, jupyterlab/jupyterlab#10370.

[balajialg]

@ryanlovett Thank you so much for circling back on the Code Mirror editor integration in Jupyter Lab. I went with the assumption that Code Mirror 6.0 was already integrated into the core interface which is not true as you clearly pointed out.

Your suggestion to wait till the CM 6.0 integration makes sense as already there are GitHub issues #1 and #2 from partner institutions which clearly documents the results from varied accessibility audit of Jupyter Lab. In high probability, we may not be adding anything new/novel to the conversation happening in this space.

However, I have a couple of questions/concerns,

• There is no proposed timeline for this integration which makes it hard to know when to scope the campus accessibility audit.
• No sign of CM 6.0 upgrade in Github issues for the retro lab (and classic notebook). Does that mean, any upgrade happening in the Jupyter Lab repo gets automatically ported into the retro interface?

I will not reach out to the accessibility audit team in the near future and wait for more clarity regarding CM 6.0 integration with retro lab & jupyter lab.

[yuvipanda]

No sign of CM 6.0 upgrade in Github issues for the retro lab (and classic notebook). Does that mean, any upgrade happening in the Jupyter Lab repo gets automatically ported into the retro interface?

Yep, anything that happens in retrolab wrt the notebook UI, text editor UI, terminals, and file browsers automatically works in retrolab.

[balajialg]

FWIW, Siteimprove gave an accessibility score of 94.3 for the datahub.berkeley.edu page.

[ryanlovett]

@balajialg That's great! It's probably very high since it didn't evaluate notebook or lab, but at least it means the landing page doesn't have a lot of problems.

[balajialg]

@ryanlovett Not sure how I missed your comment! It makes sense. As a next step, I am trying to do the Siteimprove audit for the new datahub page and add recommendations for the black antelope team to work on.

[balajialg]

Beta version (4.0.0b0) of Jupyter Lab 4.0 (JL 4.0) which is the pre-release version for Jupyter Lab released a couple of days ago. The beta version of Jupyter Notebook 7.0 (JN 7.0) is currently a work in progress at the time of writing but we have the latest alpha version of JN 7.0 (Version 7.0.0a18) which can be audited.

To track the progress of a11y changes, I am comparing the results of running the Axe Deque tool against Jupyter Lab version 3.4.5, Retro Lab version 0.3.21, and Classic Notebook version 6.4.2 which were previously installed in Data 100 hub image against the latest releases which were installed in a11y hub. Note that the current releases only have Jupyter Notebook 7.0 and Jupyter Lab 4.0 as Classic Notebook interface did not get the a11y updates and will not be supported after 2 years and Retro Lab interfaces got renamed as JN 7.0.

To make the comparison apples to apples, I used the demo notebook for lecture 2 from Data 100 course work during Sp 23 for the a11y audit.

Sharing the snapshot of the results and the spreadsheet with issues generated from running automated a11y check using Axe Deque tool for,

Classic Notebook version 6.4.2

data100.datahub.berkeley.edu-2023-03-29.csv

Here is the snapshot of the results and the spreadsheet for

RetroLab version 0.3.21

data100.datahub.berkeley.edu-2023-03-29.csv

Here is the snapshot of the results and the spreadsheet for

JupyterLab version 3.4.5

data100.datahub.berkeley.edu-2023-03-29 (1).csv

Now, sharing the snapshot of the results and the spreadsheet for the latest

Jupyterlab version 4.0.0b0

a11y.datahub.berkeley.edu-2023-03-29 (1).csv

Sharing the snapshot of the results and the spreadsheet for

Jupyter Notebook version 7.0.0a18

a11y.datahub.berkeley.edu-2023-03-29.csv

[ericvd-ucb]

Wow thanks @balajialg this is really hopeful!

[balajialg]

Keyboard test results for JL 4.0:

Keyboard test results for JN 7.0:

[balajialg]

Open issues with screen reader testing using nvda:

I installed the latest version of nvda (nvda version 2023) in my local device and allowed it to decipher a notebook from Data 8 course,

• Screen reader did not decipher text included in images without accessible text (as seen in the snapshot below),
  
• Screen Reader did not decipher the output of running a code block (as seen in the snapshot below),
  
• Screen Readers did not identify icons (as seen in the snapshot below),
  

[balajialg]

Tagging @jtpio - Feedback shared from testing beta versions of JL 4.0 and JN 7.0 using Axe Deque and Nvda in the previous comments. Hoping that you are aware of some of the issues highlighted above.

[jtpio]

Thanks @balajialg for the ping.

Also cc @brichet who has been helping with accessibility in JupyterLab and Notebook 7.

[brichet]

Thank you for sharing these tests @balajialg