Would the work being done in #24866 based off issue #24419 solve this issue?

I don't think so. If I understood #24419 and #24866 correctly, the proposed solution uses font_manager.addFont to add a font file to the manager. I've tried this and the fonts are only added for this particular session, not permanently. Which means that in every script/project I use matplotlib in, I manually have to add the fonts to the font manager using the proposed piece of code. It's not a very sustainable solution. In case there actually is no sustainable solution, I'm happy to open an issue with a feature request. I'm also happy to stand corrected in case I understood sth wrong :)

The cache file that matters is now fontlist-v330.json and

Curious. Where would this json-file appear? and how do I trigger to rebuild? In earlier versions there was font_manager.rebuild().

It should reside in ~/.matplotlib or similar.

Edit: or whereever $MPLCONFIGDIR may redirect it.

Ok, solved it. The fontlist-v330.json was in ~/.cache/matplotlib/. I deleted it, then it worked. The env $MPLCONFIGDIR was empty, though.

I'd say my initial point stands, though: the documentation does not do a particularly great job at communicating how to solve this particular problem. Since many encountered the same problem in the past, Google is filled with solutions on how to achieve loading custom fonts in older versions of matplotlib, which, in my opinion, makes it all the more important that the issue is addressed in the official documentation for current versions.

I think we need a public "please nuke the font cache and rebuild it" function in matplotlib.font_manager. That would I think solve a lot of these problems (and alleviate the need for users to understand the details of how we cache the locations and meta-data of the fonts).

I don't think it would hurt to document the cache file location as well, and/or provide a public function to expose it. Even if we don't want to support people editing this manually, it may help folks debug their font issues if they can see what is going on.

I think the cache file location is returned by matplotlib.get_cachedir(). I remember seeing it on the website.

Edit: found it. https://matplotlib.org/stable/users/faq/troubleshooting_faq.html#matplotlib-configuration-and-cache-directory-locations

Hi. I came across this issue while trying to install a dyslexic-friendly font in my matplotlib. I was following a tutorial I found, but it took me a while to figure out I had to nuke fontlist-v330.json. Anyhow, given that dyslexia is not such an uncommon characteristic in an auditorium and that the word dyslexia doesn't look to appear in matplotlib documentation... I thought of, perhaps either:

• including an example using this particular font in this bit of documentation, or
• including a dyslexic-friendly font in matptlolib's standard font library.
  The first will be easier I think, but the second may have more impact, though I have no clue how feasible that would be.
  Any thoughts on this?

I'm more inclined towards the first because we don't have the expertise to be an authority on what makes for a good dyslexic-friendly font, but I think that's a great motivation for using this feature and so would be all for including it in an example.

I'm going to send a PR for this so just to be clear, the work here is:

• Add a new "nuke/rebuild the font cache" function to the matplotlib.font_manager class.
• Write a new example using a dyslexic friendly font that shows how to add a new font file to mpl-data/ttf/ and rebuild cache. (Maybe can go in https://matplotlib.org/stable/gallery/text_labels_and_annotations/font_file.html?)
• Maybe add a link to the "expose the font cache" function in the new example: https://matplotlib.org/stable/api/matplotlib_configuration_api.html#matplotlib.get_cachedir
• Add a next_whats_new entry