I agree with this, but I forget what our policy is on assuming that users have a terminal that supports unicode.

Also, what about the other (more common) Greek letters? It would look quite weird to only have some of them changed to unicode.

At JabRef we have made good experiences with the small latex2unicode library. It's written in Scala, so not reusable here but the encoding map is rather complete: https://github.com/tomtung/latex2unicode/blob/master/src/main/scala/com/github/tomtung/latex2unicode/helper/Escape.scala

I forget what our policy is on assuming that users have a terminal that supports unicode.

I don't think we have a policy for this formulated anywhere. The sage.tensor and sage.manifolds packages already use Unicode characters, and I don't think we have heard complaints about that.

I wouldn't be concerned about terminal capabilities. I'd think the most plausible compatibility issue would be (1) workflows in which users copy-paste terminal output into a LaTeX document, without loading the necessary packages for input and font configuration; or (2) programs that run Sage as a subprocess.

At JabRef we have made good experiences with the small latex2unicode library. It's written in Scala, so not reusable here but the encoding map is rather complete: https://github.com/tomtung/latex2unicode/blob/master/src/main/scala/com/github/tomtung/latex2unicode/helper/Escape.scala

Thanks for the pointer! Looking great. I think somewhere in IPython/Jupyter there also must be a package that contains such mappings already, for offering tab-completion with the latex names.

what about the other (more common) Greek letters? It would look quite weird to only have some of them changed to unicode.

Yes, it's not complete, of course; this is just a mockup that makes CliffordAlgebra?, ExteriorAlgebra? and some tensor docstrings look good.

Also, because the substitutions are regex-based, there is a potential for breakage from unintended matches. For example, so far I have shied away from handling the \ (explicit space) command. So I think we need some kind of systematic testing to avoid unwelcome surprises.

And finally, it seems to me that this is something that should be taken care of in a more reusable way, perhaps a Sphinx extension. Mildly related:

• Replace sage.misc.sphinxify with docrepr #33682

what about the other (more common) Greek letters? It would look quite weird to only have some of them changed to unicode.

Yes, it's not complete, of course; this is just a mockup that makes CliffordAlgebra?, ExteriorAlgebra? and some tensor docstrings look good.

The unfortunate side-effect is that it makes any doc that only does part of it look very broken. This is a good proof-of-concept right now, but I think it would be hard to convince people that our doc formatter is not horribly broken without doing certain subsets (e.g., all Greek letters).

Also, because the substitutions are regex-based, there is a potential for breakage from unintended matches. For example, so far I have shied away from handling the \ (explicit space) command. So I think we need some kind of systematic testing to avoid unwelcome surprises.

Indeed, that might be a hard one to deal with. I think we would be better of replacing our docstrings with something like \quad that is easier to identify (or at least far less likely to be a misidentified).

At JabRef we have made good experiences with the small latex2unicode library. It's written in Scala, so not reusable here but the encoding map is rather complete: https://github.com/tomtung/latex2unicode/blob/master/src/main/scala/com/github/tomtung/latex2unicode/helper/Escape.scala

Thanks for the pointer! Looking great. I think somewhere in IPython/Jupyter there also must be a package that contains such mappings already, for offering tab-completion with the latex names.

I found these but haven't looked at them in any detail:

• https://pypi.org/project/latexcodec/
• https://pypi.org/project/pylatexenc/ (which took code from the previous item)
• https://github.com/mennucc/unicode2latex/blob/main/latex2unicode

Documentation preview for this PR (built with commit e17b688; changes) is ready! 🎉