From 76053fb4a9f6ec5a528fbf9232cb3e55a2cc5cc7 Mon Sep 17 00:00:00 2001
From: Eugen Rochko <eugen@zeonfederated.com>
Date: Mon, 27 Oct 2025 18:18:01 +0100
Subject: [PATCH] Fix hashtags not being picked up when full-width hash sign is
 used (#36103)

Co-authored-by: Claire <claire.github-309c@sitedethib.com>
---
 app/javascript/mastodon/actions/compose.js         | 11 ++++++-----
 .../mastodon/components/autosuggest_input.jsx      |  2 +-
 .../mastodon/components/autosuggest_textarea.jsx   |  2 +-
 app/lib/extractor.rb                               |  2 +-
 app/models/tag.rb                                  |  2 +-
 spec/lib/extractor_spec.rb                         | 14 +++++++++++++-
 spec/models/tag_spec.rb                            |  4 ++++
 7 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/app/javascript/mastodon/actions/compose.js b/app/javascript/mastodon/actions/compose.js
index ccb69f0a3d..d6de589e90 100644
--- a/app/javascript/mastodon/actions/compose.js
+++ b/app/javascript/mastodon/actions/compose.js
@@ -622,6 +622,7 @@ export function fetchComposeSuggestions(token) {
       fetchComposeSuggestionsEmojis(dispatch, getState, token);
       break;
     case '#':
+    case '＃':
       fetchComposeSuggestionsTags(dispatch, getState, token);
       break;
     default:
@@ -663,11 +664,11 @@ export function selectComposeSuggestion(position, token, suggestion, path) {
 
       dispatch(useEmoji(suggestion));
     } else if (suggestion.type === 'hashtag') {
-      completion    = `#${suggestion.name}`;
-      startPosition = position - 1;
+      completion    = suggestion.name.slice(token.length - 1);
+      startPosition = position + token.length;
     } else if (suggestion.type === 'account') {
-      completion    = getState().getIn(['accounts', suggestion.id, 'acct']);
-      startPosition = position;
+      completion    = `@${getState().getIn(['accounts', suggestion.id, 'acct'])}`;
+      startPosition = position - 1;
     }
 
     // We don't want to replace hashtags that vary only in case due to accessibility, but we need to fire off an event so that
@@ -727,7 +728,7 @@ function insertIntoTagHistory(recognizedTags, text) {
     // complicated because of new normalization rules, it's no longer just
     // a case sensitivity issue
     const names = recognizedTags.map(tag => {
-      const matches = text.match(new RegExp(`#${tag.name}`, 'i'));
+      const matches = text.match(new RegExp(`[#＃]${tag.name}`, 'i'));
 
       if (matches && matches.length > 0) {
         return matches[0].slice(1);
diff --git a/app/javascript/mastodon/components/autosuggest_input.jsx b/app/javascript/mastodon/components/autosuggest_input.jsx
index f707a18e1d..267c044215 100644
--- a/app/javascript/mastodon/components/autosuggest_input.jsx
+++ b/app/javascript/mastodon/components/autosuggest_input.jsx
@@ -61,7 +61,7 @@ export default class AutosuggestInput extends ImmutablePureComponent {
 
   static defaultProps = {
     autoFocus: true,
-    searchTokens: ['@', ':', '#'],
+    searchTokens: ['@', '＠', ':', '#', '＃'],
   };
 
   state = {
diff --git a/app/javascript/mastodon/components/autosuggest_textarea.jsx b/app/javascript/mastodon/components/autosuggest_textarea.jsx
index 68cf9e17fc..137bad9b7e 100644
--- a/app/javascript/mastodon/components/autosuggest_textarea.jsx
+++ b/app/javascript/mastodon/components/autosuggest_textarea.jsx
@@ -25,7 +25,7 @@ const textAtCursorMatchesToken = (str, caretPosition) => {
     word = str.slice(left, right + caretPosition);
   }
 
-  if (!word || word.trim().length < 3 || ['@', ':', '#'].indexOf(word[0]) === -1) {
+  if (!word || word.trim().length < 3 || ['@', '＠', ':', '#', '＃'].indexOf(word[0]) === -1) {
     return [null, null];
   }
 
diff --git a/app/lib/extractor.rb b/app/lib/extractor.rb
index 7e647a7587..206d989bf3 100644
--- a/app/lib/extractor.rb
+++ b/app/lib/extractor.rb
@@ -54,7 +54,7 @@ module Extractor
   end
 
   def extract_hashtags_with_indices(text, _options = {})
-    return [] unless text&.index('#')
+    return [] unless text&.index(/[#＃]/)
 
     possible_entries = []
 
diff --git a/app/models/tag.rb b/app/models/tag.rb
index dff1011112..1514a44005 100644
--- a/app/models/tag.rb
+++ b/app/models/tag.rb
@@ -41,7 +41,7 @@ class Tag < ApplicationRecord
   HASHTAG_LAST_SEQUENCE = '([[:word:]_]*[[:alpha:]][[:word:]_]*)'
   HASHTAG_NAME_PAT = "#{HASHTAG_FIRST_SEQUENCE}|#{HASHTAG_LAST_SEQUENCE}".freeze
 
-  HASHTAG_RE = %r{(?<![=/)\p{Alnum}])#(#{HASHTAG_NAME_PAT})}
+  HASHTAG_RE = %r{(?<![=/)\p{Alnum}])[#|＃](#{HASHTAG_NAME_PAT})}
   HASHTAG_NAME_RE = /\A(#{HASHTAG_NAME_PAT})\z/i
   HASHTAG_INVALID_CHARS_RE = /[^[:alnum:]\u0E47-\u0E4E#{HASHTAG_SEPARATORS}]/
 
diff --git a/spec/lib/extractor_spec.rb b/spec/lib/extractor_spec.rb
index bc3ee8ac49..e1a57d5788 100644
--- a/spec/lib/extractor_spec.rb
+++ b/spec/lib/extractor_spec.rb
@@ -35,12 +35,24 @@ RSpec.describe Extractor do
   end
 
   describe 'extract_hashtags_with_indices' do
-    it 'returns an empty array if it does not have #' do
+    it 'returns an empty array if it does not have # or ＃' do
       text = 'a string without hash sign'
       extracted = described_class.extract_hashtags_with_indices(text)
       expect(extracted).to eq []
     end
 
+    it 'returns hashtags preceded by an ASCII hash' do
+      text = 'hello #world'
+      extracted = described_class.extract_hashtags_with_indices(text)
+      expect(extracted).to eq [{ hashtag: 'world', indices: [6, 12] }]
+    end
+
+    it 'returns hashtags preceded by a full-width hash' do
+      text = 'hello ＃world'
+      extracted = described_class.extract_hashtags_with_indices(text)
+      expect(extracted).to eq [{ hashtag: 'world', indices: [6, 12] }]
+    end
+
     it 'does not exclude normal hash text before ://' do
       text = '#hashtag://'
       extracted = described_class.extract_hashtags_with_indices(text)
diff --git a/spec/models/tag_spec.rb b/spec/models/tag_spec.rb
index 18378c000d..d41d3a9e21 100644
--- a/spec/models/tag_spec.rb
+++ b/spec/models/tag_spec.rb
@@ -84,6 +84,10 @@ RSpec.describe Tag do
       expect(subject.match('this is #ａｅｓｔｈｅｔｉｃ').to_s).to eq '#ａｅｓｔｈｅｔｉｃ'
     end
 
+    it 'matches ＃ｆｏｏ' do
+      expect(subject.match('this is ＃ｆｏｏ').to_s).to eq '＃ｆｏｏ'
+    end
+
     it 'matches digits at the start' do
       expect(subject.match('hello #3d').to_s).to eq '#3d'
     end