I'm not sure ignoring the font properties like *TeX does is a good idea. My guess is that it's done primarily for historical reasons. This is really something that should be fixed in the font itself. There might be a good reason for why a font author would like an opinionated baseline for their symbols.

cc: @alerque

MS Word also centers the integral sign on the axis; see its results below. Presumably it's doing so motivated by the TeX math layout algorithm.

Here's MathML rendered by Firefox (same expression, same Libertinus Math font):

Again, the integral sign is centered on the axis. I dug into the MathML standard to find out why.

Symbols in MathML are represented by the <mo> element, which has properties that include stretchy, symmetric and largeop which can be set explicitly. The largeop property controls whether the glyph gets bigger in display mode (corresponding to MathClass::Large, and the symmetric property controls whether the symbol is typeset centered on the axis.

For something like <mo>∫</mo> where these properties aren't explicitly set, they are inferred from the unicode character and some additional info about its placement relative to other elements (prefix/infix/postfix). Table 25 of the Math ML Core Spec "how to determine the category of an operator" shows how characters and positions map to classes A-M of Table 26. The large operators are listed in this table only in prefix position, and appear as either class H or class J. Class H has properties symmetric and largeop and class J has properties symmetric, largeop and moveablelimits. That is, H is for integrals, and J is for all the other large operators.

The upshot is that, by default, MathML makes every large operator symmetric about the axis. This can be overridden by the content author by setting the symmetric property.

In summary: TeX's layout algorithm bakes in centering large operators on the axis (and hence its offspring do as well), MS Word appears to do the same, and it's the default behaviour in MathML. A math font designer working in this environment would know all of this, and whatever might motivate the choice of baseline for a large operator, it wouldn't be to affect default math layout because the standard of the major engines is to center the glyph on the axis. Maybe in the future Typst will allow a symmetric toggle, but that seems pretty niche. Typst's default should be to follow the other standards and center the large operator glyphs.

Oh, and this is a two-fer! Closes #1272

I'm convinced. Thank you!