Where does remark-math declare its HTML?

[benrbray]

Overview

• I am attempting to extend my remark-cite package with a rehype-cite plugin for generating HTML
• I'm using remark-math and rehype-katex as a reference for how to accomplish this
• I can't figure out where the remark-math plugin passes the mathHtml extension to unified

Details

I know that rehype-katex works by looking for math-inline and math-display elements in the hast tree:

https://github.com/remarkjs/remark-math/blob/dd3075a597d243671bf2bbb46400e3e0f3577078/packages/rehype-katex/lib/index.js#L47-L56

The math-inline and math-display classes are assigned by the mathHtml plugin exported by micromark-extension-math:

https://github.com/micromark/micromark-extension-math/blob/6fd7df3338fee9a2490cd32bee54d71c0af48581/dev/lib/html.js#L33-L46

If we have a processor defined as follows:

import remarkMath from "remark-math";

const processor = unified()
  .use(remarkParse)
  .use(remarkMath)
  .use(remarkRehype)

const markdown = "look it's $math$!";
const mdast = processor.parse(markdown);
const hast = processor.runSync(mdast);

The hast tree will contain a math-inline node:

However, the remark-math repository does not include any reference to the mathHtml plugin exported by micromark-extension-math.

https://github.com/search?q=repo%3Aremarkjs%2Fremark-math%20mathHtml&type=code

(see github search results above -- but also dug through the code manually)

Questions

• how does remark-math know how to generate HTML for math nodes if mathHtml is never imported?
• if I have a remark plugin that defines a new node type, what's the best way to allow that node type to be detected & processed by a rehype plugin?

Thank You!

[ChristianMurphy]

Hey @benrbray 👋
Went dealing with known or new node types that you want specific HTML output, leverage .data.hName, .data.hProperties, and .data.hChildren https://github.com/syntax-tree/mdast-util-to-hast?tab=readme-ov-file#fields-on-nodes

That is also how remark-math works https://github.com/syntax-tree/mdast-util-math/blob/23647664692ae59be25896d0099ba3f7157d6a76/lib/index.js#L53-L68

[wooorm]

Yup, as Christian mentions.

I’d recommend doing as little as possible in h* and such.
You have to think about: what if someone uses my syntax in plain markdown: they parse and format and serialize markdown. Will it work well? Will my markdown/remark/mdast integrations work well?
And also: what if someone writes HTML manually? Will my html/rehype/hast integrations work well for them, too?