MJ4.0.0-beta.4: \rm in subscript yields Math input error

[AlexEdgcomb]

Issue Summary

\rm in the subscript yields Math input error in beta.4, but not beta.3 and earlier

Steps to Reproduce:

1. Go to https://codepen.io/alexedgcomb/pen/BabMpmG
   Observed: \(a_\rm{b}\) yields `Math input error

Expected: No such error, like in 4.0.0-beta.3
https://codepen.io/alexedgcomb/pen/QWoYdOR

Technical details:

• MathJax Version: 4.0.0-beta.4
• Client OS: MacOS 13.4
• Browser: Firefox 122 and Chrome 121

I am not using a MathJax configuration and loading MathJax via

<script src="https://cdn.jsdelivr.net/npm/mathjax@4.0.0-beta.4/tex-mml-svg.js"></script>

Supporting information:

Interestingly, the following typeset as expected:

• \(\rm{b}\): So, not just a bad \rm
• \(a_\textrm{b}\): So, not just a bad subscript followed by command

Repro steps:

1. Go to https://codepen.io/alexedgcomb/pen/QWoYdOR
   Observed:

[dpvc]

Actually, both v3 and v4 are incorrect for this, as a_\rm{b} is not a valid LaTeX expression (actual LaTeX throws an error for this). Note that \rm is not a macro that takes an argument, like \mathrm, but is a switch that makes everything following it be in upright letters. So a_\rm{b} is effectively {a_\rm}{b}, which produces an error in LaTeX (due to how \rm is implemented). Note that if you did \rm{b}x then the x would also be upright since \rm makes everything after it be upright. The macros \bf, \it, etc. also work the same way. So you should do {\rm b} not \rm{b}, and a_{\rm b} would do what you expect.

This is in contrast to commands like \mathrm, \textrm, \symrm, and so on, which take an argument and put that in upright form. Because of that (and how they are implemented), a_\mathrm{b} is effectively the same as a_{\rm b}, which is a valid expression. Similarly, \mathrm{b}x should have the b upright, but the x in italics, where \rm{b}x has both upright.

I was surprised to see that v3 didn't give an error for a_\rm{b} as it should have, but looking into the internals, I have found out why. It is not an easy fix, and I will have to think about how best to handle it.

The error in v4 is due to new code that tries to attach to the internal MathML nodes the LaTeX that produced that node, and it is crashing on this construction. I know where it is happening, but haven't tried to resolve the issue yet.

In any case, you should be using either a_{\rm b} or a_\mathrm{b}, but not a_\rm{b}, which will produce an error in future versions.

[AlexEdgcomb]

Thank you for explaining!