FIX: remove diacritics when tokenizing html for search

app/services/search_indexer.rb

@@ -121,7 +121,8 @@ class SearchIndexer
   def self.index(obj, force: false)
     return if @disabled
 
-    category_name, tag_names = nil
+    category_name = nil
+    tag_names = nil
     topic = nil
 
     if Topic === obj
@@ -148,8 +149,7 @@ class SearchIndexer
 
     if Topic === obj && (obj.saved_change_to_title? || force)
       if obj.posts
-        post = obj.posts.find_by(post_number: 1)
+        if post = obj.posts.find_by(post_number: 1)
-        if post
           SearchIndexer.update_posts_index(post.id, obj.title, category_name, tag_names, post.cooked)
           SearchIndexer.update_topics_index(obj.id, obj.title, post.cooked)
         end
@@ -175,43 +175,22 @@ class SearchIndexer
     def self.scrub(html)
       me = new
       parser = Nokogiri::HTML::SAX::Parser.new(me)
-      begin
+      parser.parse("<div>#{html}</div>") if html.present?
-        copy = +"<div>"
-        copy << html unless html.nil?
-        copy << "</div>"
-        parser.parse(html) unless html.nil?
-      end
       me.scrubbed
     end
 
-    def start_element(name, attributes = [])
+    ATTRIBUTES ||= %w{alt title href data-youtube-title}
+
+    def start_element(_, attributes = [])
       attributes = Hash[*attributes.flatten]
-      if attributes["alt"]
+
-        scrubbed << " "
+      ATTRIBUTES.each do |name|
-        scrubbed << attributes["alt"]
+        characters(attributes[name]) if attributes[name].present?
-        scrubbed << " "
-      end
-      if attributes["title"]
-        scrubbed << " "
-        scrubbed << attributes["title"]
-        scrubbed << " "
-      end
-      if attributes["data-youtube-title"]
-        scrubbed << " "
-        scrubbed << attributes["data-youtube-title"]
-        scrubbed << " "
-      end
-      if attributes["href"]
-        scrubbed << " "
-        scrubbed << attributes["href"]
-        scrubbed << " "
       end
     end
 
     def characters(string)
-      scrubbed << " "
+      scrubbed << " #{ActiveSupport::Inflector.transliterate(string).strip} "
-      scrubbed << string
-      scrubbed << " "
     end
   end
 end

spec/services/search_indexer_spec.rb

@@ -29,6 +29,14 @@ describe SearchIndexer do
     expect(scrubbed).to eq(" http://meta.discourse.org/  link ")
   end
 
+  it 'removes diacritics' do
+    html = "<p>Hétérogénéité</p>"
+
+    scrubbed = SearchIndexer::HtmlScrubber.scrub(html)
+
+    expect(scrubbed).to eq(" Heterogeneite ")
+  end
+
   it 'correctly indexes a post according to version' do
     # Preparing so that they can be indexed to right version
     SearchIndexer.update_posts_index(post_id, "dummy", "", nil, nil)