Merge pull request #5244 from akissinger/katex

LaTeX rendering in element-web using KaTeX

Author: jryans

package.json

@@ -76,6 +76,8 @@
     "highlight.js": "^10.1.2",
     "html-entities": "^1.3.1",
     "is-ip": "^2.0.0",
+    "katex": "^0.12.0",
+    "cheerio": "^1.0.0-rc.3",
     "linkifyjs": "^2.1.9",
     "lodash": "^4.17.19",
     "matrix-js-sdk": "github:matrix-org/matrix-js-sdk#develop",

src/HtmlUtils.tsx

@@ -27,9 +27,12 @@ import _linkifyString from 'linkifyjs/string';
 import classNames from 'classnames';
 import EMOJIBASE_REGEX from 'emojibase-regex';
 import url from 'url';
+import katex from 'katex';
+import { AllHtmlEntities } from 'html-entities';
+import SettingsStore from './settings/SettingsStore';
+import cheerio from 'cheerio';
 
 import {MatrixClientPeg} from './MatrixClientPeg';
-import SettingsStore from './settings/SettingsStore';
 import {tryTransformPermalinkToLocalHref} from "./utils/permalinks/Permalinks";
 import {SHORTCODE_TO_EMOJI, getEmojiFromUnicode} from "./emoji";
 import ReplyThread from "./components/views/elements/ReplyThread";
@@ -240,7 +243,8 @@ const sanitizeHtmlParams: IExtendedSanitizeOptions = {
     allowedAttributes: {
         // custom ones first:
         font: ['color', 'data-mx-bg-color', 'data-mx-color', 'style'], // custom to matrix
-        span: ['data-mx-bg-color', 'data-mx-color', 'data-mx-spoiler', 'style'], // custom to matrix
+        span: ['data-mx-maths', 'data-mx-bg-color', 'data-mx-color', 'data-mx-spoiler', 'style'], // custom to matrix
+        div: ['data-mx-maths'],
         a: ['href', 'name', 'target', 'rel'], // remote target: custom to matrix
         img: ['src', 'width', 'height', 'alt', 'title'],
         ol: ['start'],
@@ -414,6 +418,21 @@ export function bodyToHtml(content: IContent, highlights: string[], opts: IOpts
         if (isHtmlMessage) {
             isDisplayedWithHtml = true;
             safeBody = sanitizeHtml(formattedBody, sanitizeParams);
+
+            if (SettingsStore.getValue("feature_latex_maths")) {
+                const phtml = cheerio.load(safeBody,
+                    { _useHtmlParser2: true, decodeEntities: false })
+                phtml('div, span[data-mx-maths!=""]').replaceWith(function(i, e) {
+                    return katex.renderToString(
+                        AllHtmlEntities.decode(phtml(e).attr('data-mx-maths')),
+                        {
+                            throwOnError: false,
+                            displayMode: e.name == 'div',
+                            output: "htmlAndMathml",
+                        });
+                });
+                safeBody = phtml.html();
+            }
         }
     } finally {
         delete sanitizeParams.textFilter;
@@ -515,7 +534,6 @@ export function checkBlockNode(node: Node) {
         case "H6":
         case "PRE":
         case "BLOCKQUOTE":
-        case "DIV":
         case "P":
         case "UL":
         case "OL":
@@ -528,6 +546,9 @@ export function checkBlockNode(node: Node) {
         case "TH":
         case "TD":
             return true;
+        case "DIV":
+            // don't treat math nodes as block nodes for deserializing
+            return !(node as HTMLElement).hasAttribute("data-mx-maths");
         default:
             return false;
     }

src/Markdown.js

@@ -23,13 +23,19 @@ const ALLOWED_HTML_TAGS = ['sub', 'sup', 'del', 'u'];
 const TEXT_NODES = ['text', 'softbreak', 'linebreak', 'paragraph', 'document'];
 
 function is_allowed_html_tag(node) {
+    if (node.literal != null &&
+        node.literal.match('^<((div|span) data-mx-maths="[^"]*"|\/(div|span))>$') != null) {
+        return true;
+    }
+
     // Regex won't work for tags with attrs, but we only
     // allow <del> anyway.
     const matches = /^<\/?(.*)>$/.exec(node.literal);
     if (matches && matches.length == 2) {
         const tag = matches[1];
         return ALLOWED_HTML_TAGS.indexOf(tag) > -1;
     }
+
     return false;
 }
 

src/editor/deserialize.ts

@@ -21,6 +21,7 @@ import { walkDOMDepthFirst } from "./dom";
 import { checkBlockNode } from "../HtmlUtils";
 import { getPrimaryPermalinkEntity } from "../utils/permalinks/Permalinks";
 import { PartCreator } from "./parts";
+import SdkConfig from "../SdkConfig";
 
 function parseAtRoomMentions(text: string, partCreator: PartCreator) {
     const ATROOM = "@room";
@@ -130,6 +131,23 @@ function parseElement(n: HTMLElement, partCreator: PartCreator, lastNode: HTMLEl
             }
             break;
         }
+        case "DIV":
+        case "SPAN": {
+            // math nodes are translated back into delimited latex strings
+            if (n.hasAttribute("data-mx-maths")) {
+                const delimLeft = (n.nodeName == "SPAN") ?
+                    (SdkConfig.get()['latex_maths_delims'] || {})['inline_left'] || "$" :
+                    (SdkConfig.get()['latex_maths_delims'] || {})['display_left'] || "$$";
+                const delimRight = (n.nodeName == "SPAN") ?
+                    (SdkConfig.get()['latex_maths_delims'] || {})['inline_right'] || "$" :
+                    (SdkConfig.get()['latex_maths_delims'] || {})['display_right'] || "$$";
+                const tex = n.getAttribute("data-mx-maths");
+                return partCreator.plain(delimLeft + tex + delimRight);
+            } else if (!checkDescendInto(n)) {
+                return partCreator.plain(n.textContent);
+            }
+            break;
+        }
         case "OL":
             state.listIndex.push((<HTMLOListElement>n).start || 1);
             /* falls through */

src/editor/serialize.ts

@@ -18,6 +18,10 @@ limitations under the License.
 import Markdown from '../Markdown';
 import {makeGenericPermalink} from "../utils/permalinks/Permalinks";
 import EditorModel from "./model";
+import { AllHtmlEntities } from 'html-entities';
+import SettingsStore from '../settings/SettingsStore';
+import SdkConfig from '../SdkConfig';
+import cheerio from 'cheerio';
 
 export function mdSerialize(model: EditorModel) {
     return model.parts.reduce((html, part) => {
@@ -38,10 +42,43 @@ export function mdSerialize(model: EditorModel) {
 }
 
 export function htmlSerializeIfNeeded(model: EditorModel, {forceHTML = false} = {}) {
-    const md = mdSerialize(model);
+    let md = mdSerialize(model);
+
+    if (SettingsStore.getValue("feature_latex_maths")) {
+        const displayPattern = (SdkConfig.get()['latex_maths_delims'] || {})['display_pattern'] ||
+            "\\$\\$(([^$]|\\\\\\$)*)\\$\\$";
+        const inlinePattern = (SdkConfig.get()['latex_maths_delims'] || {})['inline_pattern'] ||
+            "\\$(([^$]|\\\\\\$)*)\\$";
+
+        md = md.replace(RegExp(displayPattern, "gm"), function(m, p1) {
+            const p1e = AllHtmlEntities.encode(p1);
+            return `<div data-mx-maths="${p1e}">\n\n</div>\n\n`;
+        });
+
+        md = md.replace(RegExp(inlinePattern, "gm"), function(m, p1) {
+            const p1e = AllHtmlEntities.encode(p1);
+            return `<span data-mx-maths="${p1e}"></span>`;
+        });
+
+        // make sure div tags always start on a new line, otherwise it will confuse
+        // the markdown parser
+        md = md.replace(/(.)<div/g, function(m, p1) { return `${p1}\n<div`; });
+    }
+
     const parser = new Markdown(md);
     if (!parser.isPlainText() || forceHTML) {
-        return parser.toHTML();
+        // feed Markdown output to HTML parser
+        const phtml = cheerio.load(parser.toHTML(),
+            { _useHtmlParser2: true, decodeEntities: false })
+
+        // add fallback output for latex math, which should not be interpreted as markdown
+        phtml('div, span').each(function(i, e) {
+            const tex = phtml(e).attr('data-mx-maths')
+            if (tex) {
+                phtml(e).html(`<code>${tex}</code>`)
+            }
+        });
+        return phtml.html();
     }
     // ensure removal of escape backslashes in non-Markdown messages
     if (md.indexOf("\\") > -1) {

src/i18n/strings/en_EN.json

@@ -755,6 +755,7 @@
     "%(senderName)s: %(reaction)s": "%(senderName)s: %(reaction)s",
     "%(senderName)s: %(stickerName)s": "%(senderName)s: %(stickerName)s",
     "Change notification settings": "Change notification settings",
+    "Render LaTeX maths in messages": "Render LaTeX maths in messages",
     "Communities v2 prototypes. Requires compatible homeserver. Highly experimental - use with caution.": "Communities v2 prototypes. Requires compatible homeserver. Highly experimental - use with caution.",
     "New spinner design": "New spinner design",
     "Message Pinning": "Message Pinning",

src/settings/Settings.ts

@@ -117,6 +117,12 @@ export interface ISetting {
 }
 
 export const SETTINGS: {[setting: string]: ISetting} = {
+    "feature_latex_maths": {
+        isFeature: true,
+        displayName: _td("Render LaTeX maths in messages"),
+        supportedLevels: LEVELS_FEATURE,
+        default: false,
+    },
     "feature_communities_v2_prototypes": {
         isFeature: true,
         displayName: _td(

test/components/views/messages/TextualBody-test.js

@@ -36,6 +36,7 @@ describe("<TextualBody />", () => {
         MatrixClientPeg.matrixClient = {
             getRoom: () => mkStubRoom("room_id"),
             getAccountData: () => undefined,
+            isGuest: () => false,
         };
 
         const ev = mkEvent({
@@ -59,6 +60,7 @@ describe("<TextualBody />", () => {
         MatrixClientPeg.matrixClient = {
             getRoom: () => mkStubRoom("room_id"),
             getAccountData: () => undefined,
+            isGuest: () => false,
         };
 
         const ev = mkEvent({
@@ -83,6 +85,7 @@ describe("<TextualBody />", () => {
             MatrixClientPeg.matrixClient = {
                 getRoom: () => mkStubRoom("room_id"),
                 getAccountData: () => undefined,
+                isGuest: () => false,
             };
         });
 
@@ -135,6 +138,7 @@ describe("<TextualBody />", () => {
                 getHomeserverUrl: () => "https://my_server/",
                 on: () => undefined,
                 removeListener: () => undefined,
+                isGuest: () => false,
             };
         });
 

yarn.lock

@@ -6206,6 +6206,13 @@ jsx-ast-utils@^2.4.1:
     array-includes "^3.1.1"
     object.assign "^4.1.0"
 
+katex@^0.12.0:
+  version "0.12.0"
+  resolved "https://registry.yarnpkg.com/katex/-/katex-0.12.0.tgz#2fb1c665dbd2b043edcf8a1f5c555f46beaa0cb9"
+  integrity sha512-y+8btoc/CK70XqcHqjxiGWBOeIL8upbS0peTPXTvgrh21n1RiWWcIpSWM+4uXq+IAgNh9YYQWdc7LVDPDAEEAg==
+  dependencies:
+    commander "^2.19.0"
+
 kind-of@^3.0.2, kind-of@^3.0.3, kind-of@^3.2.0:
   version "3.2.2"
   resolved "https://registry.yarnpkg.com/kind-of/-/kind-of-3.2.2.tgz#31ea21a734bab9bbb0f32466d893aea51e4a3c64"