Don't consider textual characters to be emoji

We were using emojibase-regex to match emoji within messages. However, the docs (https://emojibase.dev/docs/regex/) state that this regex matches both emoji and text presentation characters. This is not what we want, and will result in false positives for characters like '↔' that could turn into an emoji if paired with a variation selector. Unfortunately, none of the other regexes provided by Emojibase do what we want either (milesj/emojibase#174). In the meantime, browser support for the RGI_Emoji character sequence class has made it feasible to write an emoji regex by hand, so that's what I've done.

.eslintrc.js

@@ -70,6 +70,11 @@ module.exports = {
                         name: "matrix-react-sdk/",
                         message: "Please use matrix-react-sdk/src/index instead",
                     },
+                    {
+                        name: "emojibase-regex",
+                        message:
+                            "This regex doesn't actually test for emoji. See the docs at https://emojibase.dev/docs/regex/ and prefer our own EMOJI_REGEX from HtmlUtils.",
+                    },
                 ],
                 patterns: [
                     {
@@ -138,6 +143,11 @@ module.exports = {
                         ],
                         message: "Please use matrix-js-sdk/src/matrix instead",
                     },
+                    {
+                        group: ["emojibase-regex/emoji*"],
+                        message:
+                            "This regex doesn't actually test for emoji. See the docs at https://emojibase.dev/docs/regex/ and prefer our own EMOJI_REGEX from HtmlUtils.",
+                    },
                 ],
             },
         ],

src/HtmlUtils.tsx

@@ -20,7 +20,6 @@ limitations under the License.
 import React, { LegacyRef, ReactNode } from "react";
 import sanitizeHtml from "sanitize-html";
 import classNames from "classnames";
-import EMOJIBASE_REGEX from "emojibase-regex";
 import katex from "katex";
 import { decode } from "html-entities";
 import { IContent } from "matrix-js-sdk/src/matrix";
@@ -46,10 +45,17 @@ const SURROGATE_PAIR_PATTERN = /([\ud800-\udbff])([\udc00-\udfff])/;
 const SYMBOL_PATTERN = /([\u2100-\u2bff])/;
 
 // Regex pattern for non-emoji characters that can appear in an "all-emoji" message
-// (Zero-Width Joiner, Zero-Width Space, Emoji presentation character, other whitespace)
-const EMOJI_SEPARATOR_REGEX = /[\u200D\u200B\s]|\uFE0F/g;
+// (Zero-Width Space, other whitespace)
+const EMOJI_SEPARATOR_REGEX = /[\u200B\s]/g;
 
-const BIGEMOJI_REGEX = new RegExp(`^(${EMOJIBASE_REGEX.source})+$`, "i");
+// Regex for emoji. This includes any RGI_Emoji sequence followed by an optional
+// emoji presentation VS (U+FE0F), but not those sequences that are followed by
+// a text presentation VS (U+FE0E). Technically this produces false negatives
+// for emoji followed by U+FE0E when the emoji doesn't have a text variant, but
+// in practice this doesn't matter.
+export const EMOJI_REGEX = /\p{RGI_Emoji}(?!\uFE0E)(?:(?<!\uFE0F)\uFE0F)?/v;
+
+const BIGEMOJI_REGEX = new RegExp(`^(${EMOJI_REGEX.source})+$`, "iv");
 
 /*
  * Return true if the given string contains emoji
@@ -266,7 +272,7 @@ export function formatEmojis(message: string | undefined, isHtmlMessage?: boolea
 
     const splitter = new GraphemeSplitter();
     for (const char of splitter.iterateGraphemes(message)) {
-        if (EMOJIBASE_REGEX.test(char)) {
+        if (EMOJI_REGEX.test(char)) {
             if (text) {
                 result.push(text);
                 text = "";

src/components/views/rooms/SendMessageComposer.tsx

@@ -15,7 +15,6 @@ limitations under the License.
 */
 
 import React, { createRef, KeyboardEvent, SyntheticEvent } from "react";
-import EMOJI_REGEX from "emojibase-regex";
 import {
     IContent,
     MatrixEvent,
@@ -70,6 +69,7 @@ import { doMaybeLocalRoomAction } from "../../../utils/local-room";
 import { Caret } from "../../../editor/caret";
 import { IDiff } from "../../../editor/diff";
 import { getBlobSafeMimeType } from "../../../utils/blobs";
+import { EMOJI_REGEX } from "../../../HtmlUtils";
 
 /**
  * Build the mentions information based on the editor model (and any related events):

src/editor/parts.ts

@@ -15,12 +15,11 @@ See the License for the specific language governing permissions and
 limitations under the License.
 */
 
-import EMOJIBASE_REGEX from "emojibase-regex";
 import { MatrixClient, RoomMember, Room } from "matrix-js-sdk/src/matrix";
 import GraphemeSplitter from "graphemer";
 
 import AutocompleteWrapperModel, { GetAutocompleterComponent, UpdateCallback, UpdateQuery } from "./autocomplete";
-import { unicodeToShortcode } from "../HtmlUtils";
+import { EMOJI_REGEX, unicodeToShortcode } from "../HtmlUtils";
 import * as Avatar from "../Avatar";
 import defaultDispatcher from "../dispatcher/dispatcher";
 import { Action } from "../dispatcher/actions";
@@ -198,7 +197,7 @@ abstract class BasePart {
 
 abstract class PlainBasePart extends BasePart {
     protected acceptsInsertion(chr: string, offset: number, inputType: string): boolean {
-        if (chr === "\n" || EMOJIBASE_REGEX.test(chr)) {
+        if (chr === "\n" || EMOJI_REGEX.test(chr)) {
             return false;
         }
         // when not pasting or dropping text, reject characters that should start a pill candidate
@@ -376,7 +375,7 @@ class NewlinePart extends BasePart implements IBasePart {
 
 export class EmojiPart extends BasePart implements IBasePart {
     protected acceptsInsertion(chr: string, offset: number): boolean {
-        return EMOJIBASE_REGEX.test(chr);
+        return EMOJI_REGEX.test(chr);
     }
 
     protected acceptsRemoval(position: number, chr: string): boolean {
@@ -574,7 +573,7 @@ export class PartCreator {
             case "\n":
                 return new NewlinePart();
             default:
-                if (EMOJIBASE_REGEX.test(getFirstGrapheme(input))) {
+                if (EMOJI_REGEX.test(getFirstGrapheme(input))) {
                     return new EmojiPart();
                 }
                 return new PlainPart();
@@ -652,7 +651,7 @@ export class PartCreator {
 
         const splitter = new GraphemeSplitter();
         for (const char of splitter.iterateGraphemes(text)) {
-            if (EMOJIBASE_REGEX.test(char)) {
+            if (EMOJI_REGEX.test(char)) {
                 if (plainText) {
                     parts.push(this.plain(plainText));
                     plainText = "";

test/HtmlUtils-test.tsx

@@ -107,6 +107,12 @@ describe("bodyToHtml", () => {
         expect(html).toMatchInlineSnapshot(`"<span class="mx_EventTile_searchHighlight">test</span> foo &lt;b&gt;bar"`);
     });
 
+    it("generates big emoji for emoji made of multiple characters", () => {
+        const { asFragment } = render(bodyToHtml({ body: "👨‍👩‍👧‍👦 ↔️", msgtype: "m.text" }, [], {}) as ReactElement);
+
+        expect(asFragment()).toMatchSnapshot();
+    });
+
     it("should generate big emoji for an emoji-only reply to a message", () => {
         const { asFragment } = render(
             bodyToHtml(
@@ -132,6 +138,12 @@ describe("bodyToHtml", () => {
         expect(asFragment()).toMatchSnapshot();
     });
 
+    it("does not mistake characters in text presentation mode for emoji", () => {
+        const { asFragment } = render(bodyToHtml({ body: "↔", msgtype: "m.text" }, [], {}) as ReactElement);
+
+        expect(asFragment()).toMatchSnapshot();
+    });
+
     describe("feature_latex_maths", () => {
         beforeEach(() => {
             jest.spyOn(SettingsStore, "getValue").mockImplementation((feature) => feature === "feature_latex_maths");

test/__snapshots__/HtmlUtils-test.tsx.snap

@@ -1,5 +1,16 @@
 // Jest Snapshot v1, https://goo.gl/fbAQLP
 
+exports[`bodyToHtml does not mistake characters in text presentation mode for emoji 1`] = `
+<DocumentFragment>
+  <span
+    class="mx_EventTile_body"
+    dir="auto"
+  >
+    ↔
+  </span>
+</DocumentFragment>
+`;
+
 exports[`bodyToHtml feature_latex_maths should not mangle code blocks 1`] = `"<p>hello</p><pre><code>$\\xi$</code></pre><p>world</p>"`;
 
 exports[`bodyToHtml feature_latex_maths should not mangle divs 1`] = `"<p>hello</p><div>world</div>"`;
@@ -8,6 +19,29 @@ exports[`bodyToHtml feature_latex_maths should render block katex 1`] = `"<p>hel
 
 exports[`bodyToHtml feature_latex_maths should render inline katex 1`] = `"hello <span class="katex"><span class="katex-mathml"><math xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mi>ξ</mi></mrow><annotation encoding="application/x-tex">\\xi</annotation></semantics></math></span><span class="katex-html" aria-hidden="true"><span class="base"><span class="strut" style="height:0.8889em;vertical-align:-0.1944em;"></span><span class="mord mathnormal" style="margin-right:0.04601em;">ξ</span></span></span></span> world"`;
 
+exports[`bodyToHtml generates big emoji for emoji made of multiple characters 1`] = `
+<DocumentFragment>
+  <span
+    class="mx_EventTile_body mx_EventTile_bigEmoji"
+    dir="auto"
+  >
+    <span
+      class="mx_Emoji"
+      title=":man-woman-girl-boy:"
+    >
+      👨‍👩‍👧‍👦
+    </span>
+     
+    <span
+      class="mx_Emoji"
+      title=":left_right_arrow:"
+    >
+      ↔️
+    </span>
+  </span>
+</DocumentFragment>
+`;
+
 exports[`bodyToHtml should generate big emoji for an emoji-only reply to a message 1`] = `
 <DocumentFragment>
   <span