Markdown stdlib

std/markdown parses Markdown and MDX for documentation checks, search indexes, and prose reports. It uses the runtime's content feature, which the Harn CLI includes. Lean embedders must enable that feature.

parse(source, options)#

Returns a MarkdownDocument. options.mdx defaults to false; set it to true to recognize JSX, imports, exports, and embedded expressions. Parsing doesn't evaluate JavaScript or load imported modules. Invalid MDX structure throws.

Lines start at one. Offsets count UTF-8 bytes from zero. Links include inline links, images, reference uses, and reference definitions. Code examples don't create headings or links. Repeated headings get unique anchors, including headings whose text already ends in a numbered suffix.

Prose blocks retain link labels and heading or list boundaries. They exclude frontmatter, code, images, tables, HTML comments, and bare web addresses. MDX mode also excludes import, export, and expression text. HTML extraction excludes script and style contents.

import { parse } from "std/markdown"

fn main(harness: Harness) -> nil {
  const document = parse("# Example\n\nRead the [guide](./guide.md).")
  harness.stdio.println(document.headings[0]!.anchor)
  harness.stdio.println(document.links[0]!.destination)
}

This prints example and ./guide.md.

reading_level(source, options)#

Returns a ReadingLevel, or nil when no English words remain. It uses the same parser and options as parse.

Scores estimate word and sentence length. They don't measure jargon, correctness, or how well a page explains its subject.

extract_text from std/document uses the same Markdown and HTML extraction. Pass {source_format: "markdown"} or {source_format: "html"} to select it.