zmarkdown: upgrade rehype-katex

[artragis]

use new version of rehype-katex

#295

[artragis]

@vhf, jI don't understand the root cause of test failure.

In fact I don't understand why we need to find viewBox and preserveAspectRation.

(sorry if you received a mail with french text).

[vhf]

@artragis Yep! It's related to this: rehypejs/rehype#2

Katex renders (used to render?) formulas as SVG with a viewBox attribute (and other camelCase attributes). Rehype doesn't handle these properly: they become e.g. view-box.

I wrote a hack based on the complete list of possible SVG attributes: https://github.com/zestedesavoir/zmarkdown/blob/1702c551a35e6490d01617a8258ebd989a0b8e53/packages/zmarkdown/utils/svg-hack.js

Assertions such as these:

zmarkdown/packages/zmarkdown/__tests__/server.tests.js

Lines 118 to 121 in 201c249

	expect(rendered).toContain('viewBox')
	expect(rendered).toContain('preserveAspectRatio')
	expect(rendered).not.toContain('view-box')
	expect(rendered).not.toContain('preserve-aspect-ratio')

are making sure that viewBox isn't replaced by view-box, etc.

These tests could be failing for two reasons: either Katex changed how they render things and it's not an SVG with a viewBox attr anymore or Katex rendering is broken in our project. 🤽‍♂️ 🐃

[vhf]

Yep, this comes from rehype-katex@1.2.0 vs 1.1.1, cf. https://gist.github.com/vhf/62ddb4e6967b8ca1a05d2c18572c20cd

Here is one of these SVG parts of formula, probably a symbol that got replaced by something that doesn't need SVG:

<svg width="400em" height="0.2em" viewBox="0 0 400000 200" preserveAspectRatio="xMinYMin slice">
<path d="M0 80H400000 v40H0z M0 80H400000 v40H0z">
</path>
</svg>

Please take a look at what gets rendered and if the formula looks the same with this newer version (no symbol or fraction line missing) as what currently is rendered in production, I suggest .skipping the failing tests and adding a comment with the URL of this issue.

[vhf]

#301