malice

export type RichNode =

  | { kind: 'text'; text: string }

  | { kind: 'element'; tag: string; attrs: Record<string, string>; children: RichNode[] }

  | { kind: 'mention'; user_id: string; children: RichNode[] }

  | { kind: 'spoiler'; children: RichNode[] }

  | { kind: 'emote'; mxc: string; alt: string };

// The tag allowlist from the Matrix spec's m.room.message HTML section, minus font colours.

const ALLOWED = new Set([

  'del', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'blockquote', 'p', 'a', 'ul', 'ol', 'sup', 'sub', 'li', 'b', 'i',

  'u', 'strong', 'em', 's', 'strike', 'code', 'hr', 'br', 'div', 'table', 'thead', 'tbody', 'tr', 'th', 'td',

  'caption', 'pre', 'span', 'details', 'summary', 'font'

]);

const DROPPED = new Set(['mx-reply', 'script', 'style', 'head', 'title', 'iframe', 'object', 'template']);

const MATRIX_TO_USER = /^https:\/\/matrix\.to\/#\/(@[^/?]+)/;

const SAFE_HREF = /^(https?:|mailto:)/i;

function attrsFor(el: Element, tag: string): Record<string, string> | null {

  const attrs: Record<string, string> = {};

  if (tag === 'a') {

    const href = el.getAttribute('href') ?? '';

    if (!SAFE_HREF.test(href)) return null;

    attrs['href'] = href;

  } else if (tag === 'ol') {

    const start = Number(el.getAttribute('start'));

    if (Number.isInteger(start) && start > 0) attrs['start'] = String(start);

  } else if (tag === 'code') {

    const lang = /(?:^|\s)(language-[\w+#-]+)/.exec(el.getAttribute('class') ?? '');

    if (lang) attrs['class'] = lang[1];

  }

  return attrs;

}

function convert(node: Node): RichNode[] {

  if (node.nodeType === Node.TEXT_NODE) return node.textContent ? [{ kind: 'text', text: node.textContent }] : [];

  if (node.nodeType !== Node.ELEMENT_NODE) return [];

  const el = node as Element;

  const tag = el.tagName.toLowerCase();

  if (DROPPED.has(tag)) return [];

  const children = () => [...el.childNodes].flatMap(convert);

  if (tag === 'img') {

    const src = el.getAttribute('src') ?? '';

    const alt = el.getAttribute('alt') ?? el.getAttribute('title') ?? '';

    if (src.startsWith('mxc://')) return [{ kind: 'emote', mxc: src, alt }];

    return alt ? [{ kind: 'text', text: alt }] : [];

  }

  if (tag === 'span' && el.hasAttribute('data-mx-spoiler')) return [{ kind: 'spoiler', children: children() }];

  if (!ALLOWED.has(tag)) return children();

  if (tag === 'a') {

    const user = MATRIX_TO_USER.exec(el.getAttribute('href') ?? '');

    if (user) {

      try {

        return [{ kind: 'mention', user_id: decodeURIComponent(user[1]), children: children() }];

      } catch {

        return children();

      }

    }

  }

  const attrs = attrsFor(el, tag);

  if (!attrs) return children();

  return [{ kind: 'element', tag: tag === 'font' ? 'span' : tag, attrs, children: children() }];

}

let scratch: HTMLElement | null = null;

/** Parses into one shared inert body, so no Document is allocated per message. */

function withParsedBody<T>(html: string, read: (body: HTMLElement) => T): T {

  scratch ??= document.implementation.createHTMLDocument('').body;

  scratch.innerHTML = html;

  try {

    return read(scratch);

  } finally {

    scratch.replaceChildren();

  }

}

/** A sanitised tree for a message's `formatted_body`, safe to render without `{@html}`. */

export function parseRich(html: string): RichNode[] {

  return withParsedBody(html, (body) => [...body.childNodes].flatMap(convert));

}