Allow most kinds of characters in URL query (fixes mastodon#8408) (mastodon#8447)

* Allow unicode characters in URL query strings

Fixes mastodon#8408

* Alternative approach to unicode support in urls

Adds PoC/idea to approch this problem.

Author: JMendyk

app/lib/formatter.rb

@@ -100,7 +100,7 @@ def encode(html)
   end
 
   def encode_and_link_urls(html, accounts = nil, options = {})
-    entities = Extractor.extract_entities_with_indices(html, extract_url_without_protocol: false)
+    entities = utf8_friendly_extractor(html, extract_url_without_protocol: false)
 
     if accounts.is_a?(Hash)
       options  = accounts
@@ -204,6 +204,43 @@ def rewrite(text, entities)
     result.flatten.join
   end
 
+  def utf8_friendly_extractor(text, options = {})
+    old_to_new_index = [0]
+
+    escaped = text.chars.map do |c|
+      output = c.ord.to_s(16).length > 2 ? CGI.escape(c) : c
+      old_to_new_index << old_to_new_index.last + output.length
+      output
+    end.join
+
+    # Note: I couldn't obtain list_slug with @user/list-name format
+    # for mention so this requires additional check
+    special = Extractor.extract_entities_with_indices(escaped, options).map do |extract|
+      # exactly one of :url, :hashtag, :screen_name, :cashtag keys is present
+      key = (extract.keys & [:url, :hashtag, :screen_name, :cashtag]).first
+
+      new_indices = [
+        old_to_new_index.find_index(extract[:indices].first),
+        old_to_new_index.find_index(extract[:indices].last),
+      ]
+
+      has_prefix_char = [:hashtag, :screen_name, :cashtag].include?(key)
+      value_indices = [
+        new_indices.first + (has_prefix_char ? 1 : 0), # account for #, @ or $
+        new_indices.last - 1,
+      ]
+
+      next extract.merge(
+        :indices => new_indices,
+        key => text[value_indices.first..value_indices.last]
+      )
+    end
+
+    standard = Extractor.extract_entities_with_indices(text, options)
+
+    Extractor.remove_overlapping_entities(special + standard)
+  end
+
   def link_to_url(entity, options = {})
     url        = Addressable::URI.parse(entity[:url])
     html_attrs = { target: '_blank', rel: 'nofollow noopener' }

spec/lib/formatter_spec.rb

@@ -74,10 +74,36 @@
     end
 
     context 'given a URL with a query string' do
-      let(:text) { 'https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&q=autolink' }
+      context 'with escaped unicode character' do
+        let(:text) { 'https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&q=autolink' }
 
-      it 'matches the full URL' do
-        is_expected.to include 'href="https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&q=autolink"'
+        it 'matches the full URL' do
+          is_expected.to include 'href="https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&q=autolink"'
+        end
+      end
+
+      context 'with unicode character' do
+        let(:text) { 'https://www.ruby-toolbox.com/search?utf8=✓&q=autolink' }
+
+        it 'matches the full URL' do
+          is_expected.to include 'href="https://www.ruby-toolbox.com/search?utf8=✓&q=autolink"'
+        end
+      end
+
+      context 'with unicode character at the end' do
+        let(:text) { 'https://www.ruby-toolbox.com/search?utf8=✓' }
+
+        it 'matches the full URL' do
+          is_expected.to include 'href="https://www.ruby-toolbox.com/search?utf8=✓"'
+        end
+      end
+
+      context 'with escaped and not escaped unicode characters' do
+        let(:text) { 'https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&utf81=✓&q=autolink' }
+
+        it 'preserves escaped unicode characters' do
+          is_expected.to include 'href="https://www.ruby-toolbox.com/search?utf8=%E2%9C%93&utf81=✓&q=autolink"'
+        end
       end
     end