I guess a question here is do we want to a) only allow a subset we think we know we will want (at this point the aria-label) b) only validate thetype of the ones we know about and be permissive on any other keys c) validate and only allow known aria keys?

c) is the most right, a) is probably too restrictive (we do not want to cut off someone who has a good idea) but exposes the least new API and b) is the least work, but leaves a bunch of foot cannons on the field.

so i think there is future potential in using these aria features on lower level plot objects when rendering as svg. there is more detail that can be added through aria at that level of detail, that is a distant future, but a cool goal.

since theartist._aria dict maps to html attributes the type is alwaysdict[str, str]. i don't think there is any reason to restrict aria names because y'all haven't found specific things to use them for and, at the end of the day. we know folks are gonna do some bonkers things on the web. the extent of the checking we'd have to do is that the values are something that can be cast to an html string.

i'm feeling a nice to have would be an alias specifically for aria label, to maybe encourage folks to use this feature in their blogs/docs.

hey y'all. i'm working though the the pr checklist. where are good places to add docs for this feature?

1. I'm not clear what people may want to do with this. Can you give one or two example usages?

the primary purpose of this is to encode web accessibility information (WAI-ARIA) into the artist. there is a new addition to ipython that includes alt text for images, this pull request is meant complement that so we have a place to pull the alt tags from. with the aria attribute we can include alt tech in a lot of documentation images.

2. As of now this is only carried around with the artists but not used anywhere. What is the plan for that?

these features will primarily be used with ipython, sphinx gallery, sphinx gallery to provide at aria-label/alt text for images.

3. I'm tempted to start very basic and make Artist.aria a simple attribute. That way we don't have to add all the function overhead. If people do strange things with that dict, it'll blow up some time later, but this is a very advanced-level feature that maybe doesn't need babysitting.

i'd be happy with this solution too if it is easiest. for onboarding new folks to project, i'd really likearist.set_aria_label available though. the aria label is the only part we REALLY need from this. this way we could run sprints that have folks addingplt.set_alt_test on documentation for other projects. the aria system is something that web developers do in fact do weird things with, not sure we can stop that.

I'll repeat the request for examples and a roadmap for how this will be used.

This PR is affected by a re-writing of our history to remove a large number of accidentally committed filessee discourse for details.

To recover this PR it will need be rebased onto the new default branch (main). There are several ways to accomplish this, but we recommend (assuming that you call the matplotlib/matplotlib remote"upstream"

git remote updategit checkout maingit merge --ff-only upstream/maingit checkout YOUR_BRANCHgit rebase --onto=main upstream/old_master# git rebase -i main # if you prefergit push --force-with-lease# assuming you are tracking your branch

If you do not feel comfortable doing this or need any help please reach out to any of the Matplotlib developers. We can either help you with the process or do it for you.

Thank you for your contributions to Matplotlib and sorry for the inconvenience.

At the very least this needs a rebase. Marking as stale for lack of response, and putting in draft. Feel free to move back to active...

I can offer a use case. Some projects use notebooks containing Figures as a component of their sphinx docs, using e.g.nbsphinx to convert a notebook into something sphinx can eventually render as HTML. RdTools is one example:source notebook andbuilt html. What I'd like to be able to do is specify figure-specific alt text in notebook code cells with something likefig.set_aria({'alt': 'Graph showing a time series of ...'}) (or maybe evenplt.plot(..., alt='...')?) and have it eventually show up as<img alt="Graph showing a time series of ..." ...> in the sphinx output HTML.

I think one way to get this working would be to store aria metadata on the Artist as this PR does, modify thematplotlib_inline backend to grab that metadata from theFigure and pass it along to the IPythondisplay() call so that it ends up in the notebook cell output metadata, and modifynbsphinx to pass it along tosphinx (seespatialaudio/nbsphinx#646).

Is this PR's approach considered workable? I'd be happy to submit the same patch in a fresh PR if so.

@kanderso-nrel Feel free to take the actual commit,2d28564 and start a new PR, keeping@tonyfast as an author (so ideally cherry-pick that commit and start your new PR from there). I believe that there are more things to be added (I haven't followed the discussion, but at least some roadmap ahead seems to be requested, as currently this allows users to add ARIA information, but it is not used anywhere).

From a quick look it would seem like I'd rather have the aria label being output on every artist that has that property set; if someone really wants to write tree-walking code to generate a single aria string emitted at the toplevel (an option discussed during the call) they can also temporarily erase the artist-level arias before outputting the svg.

As discussed on the call this week, we need some more expert feedback on how to make sure this is useful. Things to consider:

• just have researchers monkey patch?
• at what level (if any) do we add thing from aria -> svg elements
• should this be just on the outer most Figure or all Artists?
• is an'alt_text' attribute onFigure sufficient for our purposes?

This is not actually a blocker for#24309 as the.. plot directive already supports passing the alt text through from the rst and we need to work with sphinx-gallery to do the same in that context.

Pushed to 3.8 as we are not going to get the right people together to sort this out in the next 2 weeks.