elk/composables/content-parse.ts

361 lines
9.4 KiB
TypeScript
Raw Normal View History

2022-12-27 19:13:50 +00:00
// @unimport-disable
2022-11-21 07:14:07 +00:00
import type { Emoji } from 'masto'
import type { Node } from 'ultrahtml'
import { DOCUMENT_NODE, ELEMENT_NODE, TEXT_NODE, h, parse, render } from 'ultrahtml'
import { findAndReplaceEmojisInText } from '@iconify/utils'
import { emojiRegEx, getEmojiAttributes } from '../config/emojis'
2022-12-27 18:37:22 +00:00
2023-01-07 09:31:48 +00:00
export interface ContentParseOptions {
emojis?: Record<string, Emoji>
markdown?: boolean
replaceUnicodeEmoji?: boolean
astTransforms?: Transform[]
}
const sanitizerBasicClasses = filterClasses(/^(h-\S*|p-\S*|u-\S*|dt-\S*|e-\S*|mention|hashtag|ellipsis|invisible)$/u)
const sanitizer = sanitize({
// Allow basic elements as seen in https://github.com/mastodon/mastodon/blob/17f79082b098e05b68d6f0d38fabb3ac121879a9/lib/sanitize_ext/sanitize_config.rb
br: {},
p: {},
a: {
href: filterHref(),
class: sanitizerBasicClasses,
rel: set('nofollow noopener noreferrer'),
target: set('_blank'),
},
span: {
class: sanitizerBasicClasses,
},
// Allow elements potentially created for Markdown code blocks above
pre: {},
code: {
class: filterClasses(/^language-\w+$/),
},
})
2023-01-04 09:26:31 +00:00
const decoder = process.client ? document.createElement('textarea') : null
export function decodeHtml(text: string) {
2023-01-04 09:26:31 +00:00
if (!decoder)
// not available when SSR
return text
2022-12-05 00:30:20 +00:00
decoder.innerHTML = text
return decoder.value
}
/**
* Parse raw HTML form Mastodon server to AST,
* with interop of custom emojis and inline Markdown syntax
*/
2023-01-07 09:31:48 +00:00
export function parseMastodonHTML(
html: string,
options: ContentParseOptions = {},
) {
const {
markdown = true,
replaceUnicodeEmoji = true,
} = options
if (markdown) {
// Handle code blocks
html = html
2023-01-07 09:31:48 +00:00
.replace(/>(```|~~~)(\w*)([\s\S]+?)\1/g, (_1, _2, lang: string, raw: string) => {
const code = htmlToText(raw)
const classes = lang ? ` class="language-${lang}"` : ''
return `><pre><code${classes}>${code}</code></pre>`
})
}
// Always sanitize the raw HTML data *after* it has been modified
2023-01-07 09:31:48 +00:00
const transforms: Transform[] = [
sanitizer,
...options.astTransforms || [],
]
if (replaceUnicodeEmoji)
transforms.push(transformUnicodeEmoji)
if (markdown)
transforms.push(transformMarkdown)
transforms.push(replaceCustomEmoji(options.emojis || {}))
transforms.push(transformParagraphs)
2023-01-07 09:31:48 +00:00
return transformSync(parse(html), transforms)
}
/**
* Converts raw HTML form Mastodon server to HTML for Tiptap editor
*/
export function convertMastodonHTML(html: string, customEmojis: Record<string, Emoji> = {}) {
2023-01-07 09:31:48 +00:00
const tree = parseMastodonHTML(html, {
emojis: customEmojis,
markdown: true,
replaceUnicodeEmoji: false,
})
return render(tree)
}
2022-11-25 16:17:15 +00:00
export function htmlToText(html: string) {
const tree = parse(html)
return (tree.children as Node[]).map(n => treeToText(n)).join('').trim()
2022-11-24 03:42:03 +00:00
}
2022-11-30 06:50:47 +00:00
export function treeToText(input: Node): string {
2022-11-24 03:42:03 +00:00
let pre = ''
2022-11-25 16:17:15 +00:00
let body = ''
let post = ''
2022-11-24 03:42:03 +00:00
if (input.type === TEXT_NODE)
return decodeHtml(input.value)
2022-11-24 03:42:03 +00:00
if (input.name === 'br')
2022-11-24 03:42:03 +00:00
return '\n'
if (['p', 'pre'].includes(input.name))
2022-11-24 03:42:03 +00:00
pre = '\n'
if (input.name === 'code') {
if (input.parent?.name === 'pre') {
const lang = input.attributes.class?.replace('language-', '')
2022-11-25 19:21:53 +00:00
2022-11-30 04:50:29 +00:00
pre = `\`\`\`${lang || ''}\n`
post = '\n```'
}
else {
pre = '`'
post = '`'
}
}
else if (input.name === 'b' || input.name === 'strong') {
2022-11-30 04:50:29 +00:00
pre = '**'
post = '**'
}
else if (input.name === 'i' || input.name === 'em') {
2022-11-30 04:50:29 +00:00
pre = '*'
post = '*'
2022-11-25 16:17:15 +00:00
}
else if (input.name === 'del') {
2022-11-30 06:50:47 +00:00
pre = '~~'
post = '~~'
}
2022-11-25 16:17:15 +00:00
if ('children' in input)
body = (input.children as Node[]).map(n => treeToText(n)).join('')
2022-11-24 03:42:03 +00:00
if (input.name === 'img') {
if (input.attributes.class?.includes('custom-emoji'))
return `:${input.attributes['data-emoji-id']}:`
if (input.attributes.class?.includes('iconify-emoji'))
return input.attributes.alt
}
2022-11-30 06:50:47 +00:00
2022-11-25 16:17:15 +00:00
return pre + body + post
2022-11-24 03:42:03 +00:00
}
// A tree transform function takes an ultrahtml Node object and returns
// new content that will replace the given node in the tree.
// Returning a null removes the node from the tree.
// Strings get converted to text nodes.
// The input node's children have been transformed before the node itself
// gets transformed.
type Transform = (node: Node) => (Node | string)[] | Node | string | null
// Helpers for transforming (filtering, modifying, ...) a parsed HTML tree
// by running the given chain of transform functions one-by-one.
function transformSync(doc: Node, transforms: Transform[]) {
function visit(node: Node, transform: Transform, isRoot = false) {
if (Array.isArray(node.children)) {
const children = [] as (Node | string)[]
for (let i = 0; i < node.children.length; i++) {
const result = visit(node.children[i], transform)
if (Array.isArray(result))
children.push(...result)
else if (result)
children.push(result)
}
node.children = children.map((value) => {
if (typeof value === 'string')
return { type: TEXT_NODE, value, parent: node }
value.parent = node
return value
})
}
return isRoot ? node : transform(node)
}
for (const transform of transforms)
doc = visit(doc, transform, true) as Node
return doc
}
// A tree transform for sanitizing elements & their attributes.
type AttrSanitizers = Record<string, (value: string | undefined) => string | undefined>
function sanitize(allowedElements: Record<string, AttrSanitizers>): Transform {
return (node) => {
if (node.type !== ELEMENT_NODE)
return node
if (!Object.prototype.hasOwnProperty.call(allowedElements, node.name))
return null
const attrSanitizers = allowedElements[node.name]
const attrs = {} as Record<string, string>
for (const [name, func] of Object.entries(attrSanitizers)) {
const value = func(node.attributes[name])
if (value !== undefined)
attrs[name] = value
}
node.attributes = attrs
return node
}
}
function filterClasses(allowed: RegExp) {
return (c: string | undefined) => {
if (!c)
return undefined
return c.split(/\s/g).filter(cls => allowed.test(cls)).join(' ')
}
}
function set(value: string) {
return () => value
}
function filterHref() {
const LINK_PROTOCOLS = new Set([
'http:',
'https:',
'dat:',
'dweb:',
'ipfs:',
'ipns:',
'ssb:',
'gopher:',
'xmpp:',
'magnet:',
'gemini:',
])
return (href: string | undefined) => {
if (href === undefined)
return undefined
// Allow relative links
if (href.startsWith('/') || href.startsWith('.'))
return href
let url
try {
url = new URL(href)
}
catch (err) {
if (err instanceof TypeError)
return undefined
throw err
}
if (LINK_PROTOCOLS.has(url.protocol))
return url.toString()
return '#'
}
}
2023-01-07 09:31:48 +00:00
function transformUnicodeEmoji(node: Node) {
if (node.type !== TEXT_NODE)
return node
2023-01-07 09:31:48 +00:00
let start = 0
const matches = [] as (string | Node)[]
findAndReplaceEmojisInText(emojiRegEx, node.value, (match, result) => {
const attrs = getEmojiAttributes(match)
matches.push(result.slice(start))
matches.push(h('img', { src: attrs.src, alt: attrs.alt, class: attrs.class }))
start = result.length + match.match.length
return undefined
})
if (matches.length === 0)
return node
2023-01-07 09:31:48 +00:00
matches.push(node.value.slice(start))
return matches.filter(Boolean)
}
function replaceCustomEmoji(customEmojis: Record<string, Emoji>): Transform {
return (node) => {
if (node.type !== TEXT_NODE)
return node
const split = node.value.split(/:([\w-]+?):/g)
if (split.length === 1)
return node
return split.map((name, i) => {
if (i % 2 === 0)
return name
const emoji = customEmojis[name]
if (!emoji)
return `:${name}:`
return h('img', { 'src': emoji.url, 'alt': `:${name}:`, 'class': 'custom-emoji', 'data-emoji-id': name })
}).filter(Boolean)
}
}
2023-01-07 09:31:48 +00:00
const _markdownReplacements: [RegExp, (c: (string | Node)[]) => Node][] = [
[/\*\*\*(.*?)\*\*\*/g, c => h('b', null, [h('em', null, c)])],
[/\*\*(.*?)\*\*/g, c => h('b', null, c)],
[/\*(.*?)\*/g, c => h('em', null, c)],
[/~~(.*?)~~/g, c => h('del', null, c)],
[/`([^`]+?)`/g, c => h('code', null, c)],
]
2023-01-07 09:31:48 +00:00
function _markdownProcess(value: string) {
const results = [] as (string | Node)[]
2023-01-07 09:31:48 +00:00
let start = 0
while (true) {
let found: { match: RegExpMatchArray; replacer: (c: (string | Node)[]) => Node } | undefined
2023-01-07 09:31:48 +00:00
for (const [re, replacer] of _markdownReplacements) {
re.lastIndex = start
2023-01-07 09:31:48 +00:00
const match = re.exec(value)
if (match) {
if (!found || match.index < found.match.index!)
found = { match, replacer }
}
}
2023-01-07 09:31:48 +00:00
if (!found)
break
2023-01-07 09:31:48 +00:00
results.push(value.slice(start, found.match.index))
results.push(found.replacer(_markdownProcess(found.match[1])))
start = found.match.index! + found.match[0].length
}
2023-01-07 09:31:48 +00:00
results.push(value.slice(start))
return results.filter(Boolean)
}
function transformMarkdown(node: Node) {
if (node.type !== TEXT_NODE)
return node
return _markdownProcess(node.value)
}
function transformParagraphs(node: Node): Node | Node[] {
// For top level paragraphs, inject an empty <p> to preserve status paragraphs in our editor (except for the last one)
if (node.parent?.type === DOCUMENT_NODE && node.name === 'p' && node.parent.children.at(-1) !== node)
return [node, h('p')]
return node
}