skolplattformen-backup/lib/parseHtml.ts

import * as h2m from 'h2m'
import { htmlDecode } from 'js-htmlencode'
import { decode } from 'he'
import {
  parse, HTMLElement, TextNode,
} from 'node-html-parser'

const noChildren = [
  'strong',
  'b',
  'em',
  'i',
  'u',
  's',
]
const trimNodes = [
  ...noChildren,
  'h1',
  'h2',
  'h3',
  'h4',
  'h5',
  'h6',
  'li',
  'a',
]
const cleanText = (node: TextNode, parentType: string): TextNode => {
  const text = (parentType && trimNodes.includes(parentType.toLowerCase()))
    ? node.rawText.trim() : node.rawText
  return new TextNode(text)
}

const deepClean = (node: HTMLElement): HTMLElement => {
  const attributes = Object.entries(node.attributes)
    .map(([key, val]) => {
      if (key === 'href' && val) {
        // eslint-disable-next-line no-param-reassign
        val = val.replace(/ /g, '%20')
      }
      return `${key}="${val}"`
    })
    .join(' ')
  const cleaned = new HTMLElement(node.tagName, {}, attributes, node.parentNode)
  node.childNodes.forEach((childNode) => {
    if (childNode instanceof HTMLElement) {
      if (node.tagName && noChildren.includes(node.tagName.toLowerCase())) {
        cleaned.childNodes.push(cleanText(new TextNode(childNode.innerText), node.tagName))
      } else {
        cleaned.childNodes.push(deepClean(childNode))
      }
    } else if (childNode instanceof TextNode) {
      cleaned.childNodes.push(cleanText(childNode, node.tagName))
    }
  })
  return cleaned
}

export const clean = (html: string = ''): string => (
  deepClean(parse(decode(html))).outerHTML
)

interface Node {
  name: string
  attrs: { [key: string]: string }
  isInPreNode: boolean
  md: string
}
const converter = 'MarkdownExtra'
const overides = {
  a: (node: Node) => `[${node.md}](${node.attrs.href})`,
}

export const toMarkdown = (html: string): string => {
  const trimmed = clean(html)
  const markdown = h2m(trimmed, { overides, converter })
  const decoded = htmlDecode(markdown)
  return decoded
}