feat(math): Support MathML bevelled fractions

[Omikhleia]

Bevelled fractions (also known as skewed or diagonal fractions) are displayed with a slanted line (slash) instead of a horizontal bar.

Although this feature is still included in MathML4, it has been excluded from the MathML Core specifications.
As a result, most browsers are unlikely to support bevelled fractions...
But SILE is not a browser -- so it has the freedom to implement a broader standard! 🍰

As stated in the in-code comments, though, MathML4 does not exactly specify how to compute the layout, and the related OpenType math parameters are, at best, unclear¹ -- and not pleasant looking; moreover there's no clear distinction for display / cramped / text modes... 🌵

So I went for my own interpretation, also looking at various TeX packages (nicefrac, xfrac) for references. I do like the result:²

Oh, as a final note -- I'm going to enter the debate on the insane glyph assembly logic once again... So yes, the proposed code draws a line. A mere line, of appropriate thickness (incl. the linethickness attribute when specified) 🍭

Footnotes

1. After writing this, I found a ConTeXt manual (Hans Hagen & Mikael P. Sundqvist, Mathematics in ConTEXt, August 7, 2024 edition): SkewedFractionHorizontalGap and SkewedFractionVerticalGap "do not make sense (for us) so we do not use them." -- It's an interesting reading, by the way. ↩

2. I don't think lowering the fraction with respect to the baseline is the right thing to do. But see previous note: the mentioned ConTeXt book does it. Well, with plenty of other options etc. -- Likely beyond our own scope for now. ↩

[Omikhleia]

On my note:

the related OpenType math parameters are, at best, unclear

I just received this Building OpenType math fonts link. N.B. I don't see that comment being posted here, so maybe the author (whom I'll therefore keep anonymous) didn't want it public? But thanks for the link, fairly instructive too!

I still think it's unclear to me what the real-life expectations are: ".... so that the bottom right of the numerator as at the top left of the denominator" -- So a / b would be different from a / c, q / b or q / a .... due to bounding boxes being used? (= ascender/descender-wise?) -- Something seems ill-defined, or I can't read between the lines?

It's not as if the W3C test suite had useful examples rather than "fake" scenarios not occurring in real life... I don't think anyone would need this "dumb" example (no, I don't have to be gentle or subtle heh¹ :), but here it is anyway:

Well, it doesn't prove anything obviously (except we'd be passing that test, I guess)

Footnotes

1. I am in playful mode: A math language considering a binomial is a fraction with a 0-thick rule can claim being "presentational" indeed. Just they shouldn't have called the tag "mfrac", right? (mumbles: "20+ years of specifying things without knowing their semantics?" 😹 ) ↩