Merge branch 'master' of github.com:discourse/discourse

app/models/post.rb

@@ -38,6 +38,7 @@ class Post < ActiveRecord::Base
 
   validates_presence_of :raw, :user_id, :topic_id
   validates :raw, length: {in: SiteSetting.min_post_length..SiteSetting.max_post_length}
+  validate :raw_quality
   validate :max_mention_validator
   validate :max_images_validator
   validate :max_links_validator
@@ -68,6 +69,18 @@ class Post < ActiveRecord::Base
     self.raw.strip! if self.raw.present?
   end
 
+  def raw_quality
+
+    sentinel = TextSentinel.new(self.raw, min_entropy: SiteSetting.body_min_entropy)    
+    if sentinel.valid?
+      # It's possible the sentinel has cleaned up the title a bit
+      self.raw = sentinel.text 
+    else
+      errors.add(:raw, I18n.t(:is_invalid)) unless sentinel.valid?
+    end
+  end
+
+
   # Stop us from posting the same thing too quickly
   def unique_post_validator
     return if SiteSetting.unique_posts_mins == 0

app/models/site_setting.rb

@@ -126,6 +126,11 @@ class SiteSetting < ActiveRecord::Base
   setting(:basic_requires_read_posts, 100)
   setting(:basic_requires_time_spent_mins, 30)
 
+  # Entropy checks
+  setting(:title_min_entropy, 10)
+  setting(:body_min_entropy, 7)
+  setting(:max_word_length, 30)
+
 
   def self.call_mothership?
     self.enforce_global_nicknames? and self.discourse_org_access_key.present?

app/models/topic.rb

@@ -2,6 +2,7 @@ require_dependency 'slug'
 require_dependency 'avatar_lookup'
 require_dependency 'topic_view'
 require_dependency 'rate_limiter'
+require_dependency 'text_sentinel'
 
 class Topic < ActiveRecord::Base
   include RateLimiter::OnCreateRecord
@@ -18,13 +19,14 @@ class Topic < ActiveRecord::Base
   rate_limit :limit_topics_per_day
   rate_limit :limit_private_messages_per_day
 
+  validate :title_quality  
   validates_presence_of :title
   validates :title, length: {in: SiteSetting.min_topic_title_length..SiteSetting.max_topic_title_length}
 
   serialize :meta_data, ActiveRecord::Coders::Hstore
 
   validate :unique_title
-  validate :nuclear_option
+  
 
   belongs_to :category
   has_many :posts
@@ -113,18 +115,20 @@ class Topic < ActiveRecord::Base
     errors.add(:title, I18n.t(:has_already_been_used)) if finder.exists?
   end
 
-  # This is bad, but people are screwing us on try.discourse.org - soon we'll replace with
-  # a much more sane validation of odd characters to allow for other languages and such.
-  def nuclear_option
-    
-    # Let presence validation catch it if it's blank
-    return if title.blank?
 
-    title.each_char do |c|
-      unless (20..126).include?(c.ord)
-        errors.add(:title, I18n.t(:invalid_characters))
-        return
-      end
+  def title_quality
+    # We don't care about quality on private messages
+    return if private_message?
+
+    sentinel = TextSentinel.new(title, 
+                                min_entropy: SiteSetting.title_min_entropy, 
+                                max_word_length: SiteSetting.max_word_length,
+                                remove_interior_spaces: true)
+    if sentinel.valid?
+      # It's possible the sentinel has cleaned up the title a bit
+      self.title = sentinel.text 
+    else
+      errors.add(:title, I18n.t(:is_invalid)) unless sentinel.valid?
     end
   end