Making tables that can be tagged

[sbreitbart-NOAA]

Over the past few months, @Schiano-NOAA and I have done extensive testing to figure out how to create asar-based reports that pass Section 508 accessibility criteria. One of the major roadblocks? Tables that could be tagged.

🎉 The good news is that, as of asar version 1.6.0., asar reports meet the NOAA "Big 5" Section 508 compliance requirements- and offer even more accessibility features!

There is a limitation in Pandoc when processing raw LaTeX blocks, which bypass tagpdf. Any tables produced by a package that follows this process, such as flextable, will not be tagged with asar::add_accessibility(), thus not meeting Section 508 standards. As we've found success with other packages like gt and kableExtra, we are changing the basis of stockplotr tables from flextable to gt. We will share updates on our progress towards making accessible reports as the development of our workflow, and our essential dependencies, evolves.

Raw LaTeX tables in Quarto pose the least issues when compiling asar reports. Next best are gt and kableExtra, though the Adobe Accessibility Check flags for issues with headers (which can be resolved). Please check out the "a11y-sandbox" repository README for findings related to differences in tagging behavior for tables produced with these packages, how to fix the header issue, and a summary of testing findings. If you care to dig deeper, the repository contains many examples of features and table types that were tested for success in tagging, passing Adobe Accessibility Checker tests, and more.

If you have trouble creating custom tables, we'd be happy to help!

[kellijohnson-NOAA]

@davidgohel first thank you for making {flextable}, we really enjoy using it to produce tables within our quarto documents. Unfortunately, we have come to a roadblock that stops us from continuing to use your excellent package because we cannot make the resulting tables accessible to those with disabilities when we produce pdf documents. We want to bring this observation to your attention in case you have time to talk about it. We are happy to provide some reproducible examples if you wish.

[davidgohel]

Your colleague already publicly recommended to "avoid flextable" before you posted this. The polite tone doesn't change what this is: a form of pressure or an insult

It's a well-known pattern in these days!

[davidgohel]

The PDF tagging issue is not a flextable problem, it's a limitation of how Pandoc processes raw LaTeX blocks, which bypasses tagpdf. Any package injecting raw LaTeX through Pandoc will have the exact same issue.

Blaming flextable for this is incorrect. You clearly don't understand where the problem lies in the rendering pipeline.

Publicly recommending to "avoid flextable" without even opening an issue first is disrespectful to the work behind this package. You could have opened an issue. Instead you chose to discourage people from using it based on YOUR misunderstanding of where the problem actually lies.

This is not the first time NOAA pushes to drop flextable. Back in 2022, your colleagues were already encouraging migration to gt (see quarto-dev/quarto-cli#1556). You never open issues, you never contribute, you just publicly recommend against using the package. That's not how open-source collaboration works.

If flextable doesn't fit your needs, stop using it. But stop publicly discrediting a package that has been actively maintained and stable for over 10 years.

[davidgohel]

Interesting that stockplotr still lists flextable as a hard dependency while NOAA publicly recommending to avoid it.

https://github.com/nmfs-ost/stockplotr/blob/main/DESCRIPTION#L41

[sbreitbart-NOAA]

@davidgohel Thank you for your clarifications regarding what part of the process in creating tables is causing the lack of tagging. We edited the wording in the main body of the discussion describing the limitation. Our intention was not to tell people to never use it, but more that we were working through the limitation, which is why we tagged you. We hope that you can see we had positive rather than negative intent. We would have opened an issue if we knew what the issue actually was but, as you point out, we did not. By tagging you, we were hopeful that you would be willing to guide us towards a positive path of helping either fix an issue or showing us mistakes we were making. By no means did we want to discredit the hard work that has been put into the package.

flextable is still a dependency in the package because we don’t expect or want everyone to convert to a different package or change their workflow. We have a function that applies formatting to flextables called add_theme() because we anticipate users are still using the package. We just needed a quick fix while we worked through the issues we were facing.

Quarto just released a new version (v1.9) that improves the accessibility of their documents. We are not sure if they were able to address the Pandoc issue. If so, that would be great! We have had success in working with them to get changes into their workflow, so if the issue still has not been addressed, we are happy to help where we can.

-Sophie, Sam, and Kelli

[Schiano-NOAA]

As of March 24, Quarto v1.9 has improved accessibility features produced with our help. I am planning on running a series of tests (as I did before) to determine the usability of a variety of table-making packages using Quarto. Our priority is making sure that the documents produced from our tools can meet most (if not all) of the federal accessibility requirements (described in Section 508).