[v3 SVG] .MathJax path

[pkra]

Just a random note after trying to reduce the stylesheet generated by the SVG output. At

https://github.com/mathjax/MathJax-src/blob/fbd71f3f7ca06dc71304e8a2c03f13be233ccac5/ts/output/svg/Wrappers/TextNode.ts#L49-L53

the stylesheet sets .MathJax path { stroke-width: 3}.

I believe this has no effect (except when disabling fontCache) since it will not match the path in a use element shadow DOM.

[dpvc]

I ran into this recently myself. It turns out that some browsers do match it and others don't. Firefox doesn't, but WebKit does, so it affects Safari and Chrome (and presumably Opera, and Blink-based Edge, though I didn't verify that), but doesn't affect Firefox. I'm not sure which is correct, but because the shadow DOM created by <use> is internal, I would expect the <use> to honor the styling of the original <path> (which is not in shadowDOM).

In any case, you are right that it doesn't affect Firefox output. Adding the same CSS for .MathJax use resolves the issue for Firefox.

[pkra]

I ran into this recently myself. It turns out that some browsers do match it and others don't.

Ah, I didn't test carefully enough. I can confirm that Chrome honors it (though it doesn't show up in the dev tools styles list).

I'm not sure which is correct, but because the shadow DOM created by <use> is internal, I would expect the <use> to honor the styling of the original <path> (which is not in shadowDOM).

I think the SVG 2 spec agrees with you, https://svgwg.org/svg2-draft/struct.html#UseStyleInheritance

So this is then really just a browser bug.

(FWIW, I can't spot a difference to the default stroke-width of 1 on any of my displays.)

[dpvc]

Yes, there is almost no difference between 1 and 3 when the em-size is 1000. The non-zero amount was to make some very thin areas (like in the italic lower-case "o") less likely to disappear, but I would guess 1 would do just as well. The idea was to have at least some stroke-width, and not just fill, since some areas get very thin.

[dpvc]

Thanks for the pointer to the documentation. I've had a quick look, and I think Firefox may be doing the right thing in this case. Apparently "complex selectors" (ones that include parent elements, like our .MathJax path are not supposed to be transferred to the shadow DOM.

In any case, assigning .MathJax use as well as .MathJax path should take care of both situations.

[pkra]

Since I just stumbled upon this again, the naming of this rule still seems odd compared to other style rules.

[dpvc]

I have made a PR to use better selectors that are more specific and target both path and use elements.