Refactor regular expression used to find tags so that it supports unicode in both ruby 1.8 and ruby 1.9. NEVER SPEAK OF THIS AGAIN.

app/models/acts_as_taggable_on_tag.rb

@@ -1,8 +1,13 @@
 class ActsAsTaggableOn::Tag
+
   def followed_count
    @followed_count ||= TagFollowing.where(:tag_id => self.id).count
   end
 
+  def self.tag_text_regexp
+    @@tag_text_regexp ||= (RUBY_VERSION.include?('1.9') ? "[[:alnum:]]_-" : "\\w-")
+  end
+
   def self.autocomplete(name)
     where("name LIKE ?", "#{name.downcase}%")
   end
@@ -12,7 +17,7 @@ class ActsAsTaggableOn::Tag
       # Special case for love, because the world needs more love.
       '<3'
     elsif name
-      name.gsub(/[^\w-]/, '')
+      name.gsub(/[^#{self.tag_text_regexp}]/, '')
     end
   end
 end

lib/diaspora/taggable.rb

@@ -25,7 +25,7 @@ module Diaspora
     end
 
     def tag_strings
-      regex = /(?:^|\s)#([\w-]+|<3)/
+      regex = /(?:^|\s)#([#{ActsAsTaggableOn::Tag.tag_text_regexp}]+|<3)/u
       matches = self.
         send( self.class.field_with_tags ).
         scan(regex).
@@ -41,7 +41,7 @@ module Diaspora
       return text if opts[:plain_text]
 
       text = ERB::Util.h(text) unless opts[:no_escape]
-      regex = /(^|\s|>)#([\w-]+|&lt;3)/
+      regex =/(^|\s|>)#([#{ActsAsTaggableOn::Tag.tag_text_regexp}]+|&lt;3)/u
 
       text.to_str.gsub(regex) { |matched_string|
         pre, url_bit, clickable = $1, $2, "##{$2}"

spec/shared_behaviors/taggable.rb

@@ -24,8 +24,15 @@ describe Diaspora::Taggable do
         @object.save!
       end
 
-      it 'links the tag to /p' do
-        Diaspora::Taggable.format_tags(@str).should include( tag_link('vöglein') )
+      it "supports non-ascii characters" do
+        @object.tags(true).map(&:name).should include('vöglein')
+      end
+
+      it 'links each tag' do
+        formatted_string = Diaspora::Taggable.format_tags(@str)
+        formatted_string.should include(tag_link('what'))
+        formatted_string.should include(tag_link('hey'))
+        formatted_string.should include(tag_link('vöglein'))
       end
 
       it 'responds to plain_text' do