MathML mathvariant support is incomplete

[Omikhleia]

Our implementation of thee MathML mathvariant attribute only supports but a few of the possibilities from MathML3.[^1]

It's true that MathML4 and MathML Core both encourage the direct use of Unicode characters, making mathvariant somewhat legacy... but the logic is useful for parity with LaTeX in our TeX math like syntax...

• [x] Support at least most of the mnemonics ("fraktur", "script", etc.)...
• [x] Support numbers and greek when applicable (and not only capital and small letters)...[^2]
• [x] Ensure TeX-like \Gamma etc. are upright (see comment below)
• [x] In TeX-like math, simple macros for \mathcal, \mathfrak and \mathbb are then within reach... (on the principle of least astonishment, and for ease of integrating TeX-based formulas from other sources).

--> I'm on this one.

Fonts such as Libertinus Serif have a good coverage for these characters, so a show case demo even becomes easy:

• Checking the mappings as I add them (checking the non-normative but really nice MathML Core Appendic C)...
• Generating a "ptable" generated by looping on variants and characters...

One always need a nice target, bits by bits... Eventually... Say the Schwinger–Dyson equation

\left\langle\psi\left|\mathcal{T}\left\{\frac{\delta}{\delta\phi}F[\phi]\right\}\right|\psi\right\rangle = -\mathrm{i}\left\langle\psi\left|\mathcal{T}\left\{F[\phi]\frac{\delta}{\delta\phi}S[\phi]\right\}\right|\psi\right\rangle`

Should give what's expected...

(Firefox rendered MathML from Pandoc's output for Wikipedia's TeX-based formula).

We are not yet there (oh those \left and \right... on another day), but there's a way. Don't you think oh my dear "math friends"? -- or here, rather, my "theoretical physicist friends" :clown_face:

[^1]: See our of mathVariantToScriptType and related code: 5 out of 18 (we might not support all of them, see next note, but we can support most. [^2]: There's some arabic stuff too, but I'll likely suggest to keep it out of this issue for now.

[Omikhleia]

Interesting side effect... Before:

After:

• This seems consistent with MathML (single characters --> italic)...
• The same formula converted by Pandoc to MathML and visualized in Firefox is rendered with italics too on the Gamma and zeta...
• But TeX seems to always have these "upright" (for capital greek letters obtained by a macro) though there are packages to change that, this seems to be the usual way)... We can obtain this by explicit macros (\def{Gamma}{\mi[mathvariant=normal]{Γ}} etc. but I wonder what the expectations are.

EDIT: MathJax has the capital greek macros in the mathchar7 parsing method (= defaults to "normal") and the lowercase greek macro in mathchar0mi (= defaults to italic). So it would seem the expectation is:

(\Gamma upright, but \zeta italic)