import {isValid} from "date-fns"; import katex from "katex"; import _ from "lodash"; import * as emoji from "../shared/js/emoji"; import * as fenced_code from "../shared/js/fenced_code"; import marked from "../third/marked/lib/marked"; import * as blueslip from "./blueslip"; import * as linkifiers from "./linkifiers"; import * as message_store from "./message_store"; // This contains zulip's frontend Markdown implementation; see // docs/subsystems/markdown.md for docs on our Markdown syntax. The other // main piece in rendering Markdown client-side is // static/third/marked/lib/marked.js, which we have significantly // modified from the original implementation. // Docs: https://zulip.readthedocs.io/en/latest/subsystems/markdown.html // This should be initialized with a struct // similar to markdown_config.get_helpers(). // See the call to markdown.initialize() in ui_init // for example usage. let helpers; // Regexes that match some of our common backend-only Markdown syntax const backend_only_markdown_re = [ // Inline image previews, check for contiguous chars ending in image suffix // To keep the below regexes simple, split them out for the end-of-message case /\S*(?:\.bmp|\.gif|\.jpg|\.jpeg|\.png|\.webp)\)?\s+/m, /\S*(?:\.bmp|\.gif|\.jpg|\.jpeg|\.png|\.webp)\)?$/m, // Twitter and youtube links are given previews /\S*(?:twitter|youtube).com\/\S*/, ]; export function translate_emoticons_to_names(text) { // Translates emoticons in a string to their colon syntax. let translated = text; let replacement_text; const terminal_symbols = ",.;?!()[] \"'\n\t"; // From composebox_typeahead const symbols_except_space = terminal_symbols.replace(" ", ""); const emoticon_replacer = function (match, g1, offset, str) { const prev_char = str[offset - 1]; const next_char = str[offset + match.length]; const symbol_at_start = terminal_symbols.includes(prev_char); const symbol_at_end = terminal_symbols.includes(next_char); const non_space_at_start = symbols_except_space.includes(prev_char); const non_space_at_end = symbols_except_space.includes(next_char); const valid_start = symbol_at_start || offset === 0; const valid_end = symbol_at_end || offset === str.length - match.length; if (non_space_at_start && non_space_at_end) { // Hello!:)? return match; } if (valid_start && valid_end) { return replacement_text; } return match; }; for (const translation of emoji.get_emoticon_translations()) { // We can't pass replacement_text directly into // emoticon_replacer, because emoticon_replacer is // a callback for `replace()`. Instead we just mutate // the `replacement_text` that the function closes on. replacement_text = translation.replacement_text; translated = translated.replace(translation.regex, emoticon_replacer); } return translated; } export function contains_backend_only_syntax(content) { // Try to guess whether or not a message contains syntax that only the // backend Markdown processor can correctly handle. // If it doesn't, we can immediately render it client-side for local echo. const markedup = backend_only_markdown_re.find((re) => re.test(content)); // If a linkifier doesn't start with some specified characters // then don't render it locally. It is workaround for the fact that // javascript regex doesn't support lookbehind. const linkifier_list = linkifiers.linkifier_list; const false_linkifier_match = linkifier_list.find((re) => { const pattern = /[^\s"'(,:<]/.source + re.pattern.source + /(?!\w)/.source; const regex = new RegExp(pattern); return regex.test(content); }); return markedup !== undefined || false_linkifier_match !== undefined; } export function apply_markdown(message) { message_store.init_booleans(message); const options = { userMentionHandler(mention, silently) { if (mention === "all" || mention === "everyone" || mention === "stream") { let classes; let display_text; if (silently) { classes = "user-mention silent"; display_text = mention; } else { message.mentioned = true; display_text = "@" + mention; classes = "user-mention"; } return `${_.escape(display_text)}`; } let full_name; let user_id; const id_regex = /^(.+)?\|(\d+)$/; // For @**user|id** and @**|id** syntax const match = id_regex.exec(mention); if (match) { /* If we have two users named Alice, we want users to provide mentions like this: alice|42 alice|99 The autocomplete feature will help users send correct mentions for duplicate names, but we also have to consider the possibility that the user will hand-type something incorrectly, in which case we'll fall through to the other code (which may be a misfeature). */ full_name = match[1]; user_id = Number.parseInt(match[2], 10); if (full_name === undefined) { // For @**|id** syntax if (!helpers.is_valid_user_id(user_id)) { // silently ignore invalid user id. user_id = undefined; } else { full_name = helpers.get_actual_name_from_user_id(user_id); } } else { // For @**user|id** syntax if (!helpers.is_valid_full_name_and_user_id(full_name, user_id)) { user_id = undefined; full_name = undefined; } } } if (user_id === undefined) { // Handle normal syntax full_name = mention; user_id = helpers.get_user_id_from_name(full_name); } if (user_id === undefined) { // This is nothing to be concerned about--the users // are allowed to hand-type mentions and they may // have had a typo in the name. return undefined; } // HAPPY PATH! Note that we not only need to return the // appropriate HTML snippet here; we also want to update // flags on the message itself that get used by the message // view code and possibly our filtering code. // If I mention "@aLiCe sMITH", I still want "Alice Smith" to // show in the pill. let display_text = helpers.get_actual_name_from_user_id(user_id); let classes; if (silently) { classes = "user-mention silent"; } else { if (helpers.my_user_id() === user_id) { message.mentioned = true; message.mentioned_me_directly = true; } classes = "user-mention"; display_text = "@" + display_text; } return `${_.escape( display_text, )}`; }, groupMentionHandler(name, silently) { const group = helpers.get_user_group_from_name(name); if (group !== undefined) { let display_text; let classes; if (silently) { display_text = group.name; classes = "user-group-mention silent"; } else { display_text = "@" + group.name; classes = "user-group-mention"; if (helpers.is_member_of_user_group(group.id, helpers.my_user_id())) { message.mentioned = true; } } return `${_.escape(display_text)}`; } return undefined; }, silencedMentionHandler(quote) { // Silence quoted mentions. quote = quote.replace( /()@/g, "$1 silent$2", ); // Silence quoted user group mentions. quote = quote.replace( /()@/g, "$1 silent$2", ); // In most cases, if you are being mentioned in the message you're quoting, you wouldn't // mention yourself outside of the blockquote (and, above it). If that you do that, the // following mentioned status is false; the backend rendering is authoritative and the // only side effect is the lack red flash on immediately sending the message. message.mentioned = false; message.mentioned_me_directly = false; return quote; }, }; // Our Python-Markdown processor appends two \n\n to input message.content = marked(message.raw_content + "\n\n", options).trim(); message.is_me_message = is_status_message(message.raw_content); } export function add_topic_links(message) { if (message.type !== "stream") { message.topic_links = []; return; } const topic = message.topic; const links = []; const linkifier_list = linkifiers.linkifier_list; for (const linkifier of linkifier_list) { const pattern = linkifier.pattern; const url = linkifier.url_format; let match; while ((match = pattern.exec(topic)) !== null) { let link_url = url; const matched_groups = match.slice(1); let i = 0; while (i < matched_groups.length) { const matched_group = matched_groups[i]; const current_group = i + 1; const back_ref = "\\" + current_group; link_url = link_url.replace(back_ref, matched_group); i += 1; } // We store the starting index as well, to sort the order of occurrence of the links // in the topic, similar to the logic implemented in zerver/lib/markdown/__init__.py links.push({url: link_url, text: match[0], index: topic.indexOf(match[0])}); } } // Also make raw URLs navigable const url_re = /\b(https?:\/\/[^\s<]+[^\s"'),.:;<\]])/g; // Slightly modified from third/marked.js const matches = topic.match(url_re); if (matches) { for (const match of matches) { links.push({url: match, text: match, index: topic.indexOf(match)}); } } links.sort((a, b) => a.index - b.index); for (const match of links) { delete match.index; } message.topic_links = links; } export function is_status_message(raw_content) { return raw_content.startsWith("/me "); } function make_emoji_span(codepoint, title, alt_text) { return `${_.escape(alt_text)}`; } function handleUnicodeEmoji(unicode_emoji) { const codepoint = unicode_emoji.codePointAt(0).toString(16); const emoji_name = emoji.get_emoji_name(codepoint); if (emoji_name) { const alt_text = ":" + emoji_name + ":"; const title = emoji_name.replace(/_/g, " "); return make_emoji_span(codepoint, title, alt_text); } return unicode_emoji; } function handleEmoji(emoji_name) { const alt_text = ":" + emoji_name + ":"; const title = emoji_name.replace(/_/g, " "); // Zulip supports both standard/Unicode emoji, served by a // spritesheet and custom realm-specific emoji (served by URL). // We first check if this is a realm emoji, and if so, render it. // // Otherwise we'll look at Unicode emoji to render with an emoji // span using the spritesheet; and if it isn't one of those // either, we pass through the plain text syntax unmodified. const emoji_url = emoji.get_realm_emoji_url(emoji_name); if (emoji_url) { return `${_.escape(alt_text)}`; } const codepoint = emoji.get_emoji_codepoint(emoji_name); if (codepoint) { return make_emoji_span(codepoint, title, alt_text); } return alt_text; } function handleTimestamp(time) { let timeobject; if (Number.isNaN(Number(time))) { timeobject = new Date(time); // not a Unix timestamp } else { // JavaScript dates are in milliseconds, Unix timestamps are in seconds timeobject = new Date(time * 1000); } const escaped_time = _.escape(time); if (!isValid(timeobject)) { // Unsupported time format: rerender accordingly. // We do not show an error on these formats in local echo because // there is a chance that the server would interpret it successfully // and if it does, the jumping from the error message to a rendered // timestamp doesn't look good. return `${escaped_time}`; } // Use html5