FEATURE: add censored_pattern setting to censor posts using regex

app/assets/javascripts/pretty-text/censored-words.js.es6

@@ -1,19 +1,38 @@
-export function censor(text, censoredWords) {
-  if (censoredWords && censoredWords.length) {
-    const split = censoredWords.split("|");
-    let censorRegexp;
-    if (split && split.length) {
-      censorRegexp = new RegExp("(\\b(?:" + split.map(function (t) { return "(" + t.replace(/[-/\\^$*+?.()|[\]{}]/g, '\\$&') + ")"; }).join("|") + ")\\b)(?![^\\(]*\\))", "ig");
-    }
+export function censor(text, censoredWords, censoredPattern) {
+  let patterns = [],
+      originalText = text;
 
-    if (censorRegexp) {
-      let m = censorRegexp.exec(text);
-      while (m && m[0]) {
-        const replacement = new Array(m[0].length+1).join('■');
-        text = text.replace(new RegExp("(\\b" + m[0] + "\\b)(?![^\\(]*\\))", "ig"), replacement);
-        m = censorRegexp.exec(text);
+  if (censoredWords && censoredWords.length) {
+    patterns = censoredWords.split("|").map(t => { return "(" + t.replace(/[-/\\^$*+?.()|[\]{}]/g, '\\$&') + ")"; });
+  }
+
+  if (censoredPattern && censoredPattern.length > 0) {
+    try {
+      new RegExp(censoredPattern); // exception if invalid
+      patterns.push("(" + censoredPattern + ")");
+    } catch(e) {}
+  }
+
+  if (patterns.length) {
+    let censorRegexp;
+
+    try {
+      censorRegexp = new RegExp("(\\b(?:" + patterns.join("|") + ")\\b)(?![^\\(]*\\))", "ig");
+
+      if (censorRegexp) {
+        let m = censorRegexp.exec(text);
+
+        while (m && m[0]) {
+          if (m[0].length > originalText.length) { return originalText; } // regex is dangerous
+          const replacement = new Array(m[0].length+1).join('■');
+          text = text.replace(new RegExp("(\\b" + m[0] + "\\b)(?![^\\(]*\\))", "ig"), replacement);
+          m = censorRegexp.exec(text);
+        }
       }
+    } catch(e) {
+      return originalText;
     }
   }
+
   return text;
 }

app/assets/javascripts/pretty-text/engines/discourse-markdown/censored.js.es6

@@ -4,10 +4,12 @@ import { registerOption } from 'pretty-text/pretty-text';
 registerOption((siteSettings, opts) => {
   opts.features.censored = true;
   opts.censoredWords = siteSettings.censored_words;
+  opts.censoredPattern = siteSettings.censored_pattern;
 });
 
 export function setup(helper) {
   helper.addPreProcessor(text => {
-    return censor(text, helper.getOptions().censoredWords);
+    const options = helper.getOptions();
+    return censor(text, options.censoredWords, options.censoredPattern);
   });
 }

config/locales/server.en.yml

@@ -828,6 +828,7 @@ en:
 
   site_settings:
     censored_words: "Words that will be automatically replaced with ■■■■"
+    censored_pattern: "Regex pattern that will be automatically replaced with ■■■■"
     delete_old_hidden_posts: "Auto-delete any hidden posts that stay hidden for more than 30 days."
     default_locale: "The default language of this Discourse instance (ISO 639-1 Code)"
     allow_user_locale: "Allow users to choose their own language interface preference"
@@ -1442,6 +1443,7 @@ en:
       reply_by_email_address_is_empty: "You must set a 'reply by email address' before enabling reply by email."
       email_polling_disabled: "You must enable either manual or POP3 polling before enabling reply by email."
       user_locale_not_enabled: "You must first enable 'allow user locale' before enabling this setting."
+      invalid_regex: "Regex is invalid or not allowed."
 
   search:
     within_post: "#%{post_number} by %{username}"

config/site_settings.yml

@@ -536,6 +536,11 @@ posting:
     default: ''
     refresh: true
     type: list
+  censored_pattern:
+    client: true
+    default: ''
+    refresh: true
+    type: regex
   enable_emoji:
     default: true
     client: true

lib/pretty_text.rb

@@ -212,7 +212,18 @@ module PrettyText
     options[:topicId] = opts[:topic_id]
 
     working_text = text.dup
-    sanitized = markdown(working_text, options)
+
+    begin
+      sanitized = markdown(working_text, options)
+    rescue MiniRacer::ScriptTerminatedError => e
+      if SiteSetting.censored_pattern.present?
+        Rails.logger.warn "Post cooking timed out. Clearing the censored_pattern setting and retrying."
+        SiteSetting.censored_pattern = nil
+        sanitized = markdown(working_text, options)
+      else
+        raise e
+      end
+    end
 
     doc = Nokogiri::HTML.fragment(sanitized)
 

lib/site_setting_extension.rb

@@ -32,7 +32,8 @@ module SiteSettingExtension
                         url_list: 9,
                         host_list: 10,
                         category_list: 11,
-                        value_list: 12)
+                        value_list: 12,
+                        regex: 13)
   end
 
   def mutex
@@ -443,7 +444,8 @@ module SiteSettingExtension
       types[:fixnum] => IntegerSettingValidator,
       types[:string] => StringSettingValidator,
       'list' => StringSettingValidator,
-      'enum' => StringSettingValidator
+      'enum' => StringSettingValidator,
+      'regex' => RegexSettingValidator
     }
     @validator_mapping[type_name]
   end

lib/validators/regex_setting_validator.rb

@@ -0,0 +1,25 @@
+class RegexSettingValidator
+
+  LOREM = 'Lorem ipsum dolor sit amet, consectetur adipiscing elit. Nullam eget sem non elit tincidunt rhoncus.'.freeze
+
+  def initialize(opts={})
+    @opts = opts
+  end
+
+  def valid_value?(val)
+    !val.present? || valid_regex?(val)
+  end
+
+  # Check that string is a valid regex, and that it doesn't match most of the lorem string.
+  def valid_regex?(val)
+    r = Regexp.new(val)
+    matches = r.match(LOREM)
+    matches.nil? || matches[0].length < (LOREM.length - 10)
+  rescue
+    false
+  end
+
+  def error_message
+    I18n.t('site_settings.errors.invalid_regex')
+  end
+end

spec/components/pretty_text_spec.rb

@@ -438,4 +438,13 @@ HTML
     end
   end
 
+  describe "censored_pattern site setting" do
+    it "can be cleared if it causes cooking to timeout" do
+      SiteSetting.censored_pattern = "evilregex"
+      described_class.stubs(:markdown).raises(MiniRacer::ScriptTerminatedError)
+      PrettyText.cook("Protect against it plz.") rescue nil
+      expect(SiteSetting.censored_pattern).to be_blank
+    end
+  end
+
 end

spec/components/validators/regex_setting_validator_spec.rb

@@ -0,0 +1,24 @@
+require 'rails_helper'
+
+describe RegexSettingValidator do
+  describe '#valid_value?' do
+    subject(:validator) { described_class.new }
+
+    it "returns true for blank values" do
+      expect(validator.valid_value?('')).to eq(true)
+      expect(validator.valid_value?(nil)).to eq(true)
+    end
+
+    it "return false for invalid regex" do
+      expect(validator.valid_value?('(()')).to eq(false)
+    end
+
+    it "returns false for regex with dangerous matches" do
+      expect(validator.valid_value?('(.)*')).to eq(false)
+    end
+
+    it "returns true for safe regex" do
+      expect(validator.valid_value?('\d{3}-\d{4}')).to eq(true)
+    end
+  end
+end

test/javascripts/lib/pretty-text-test.js.es6

@@ -11,7 +11,8 @@ const defaultOpts = buildOptions({
     emoji_set: 'emoji_one',
     highlighted_languages: 'json|ruby|javascript',
     default_code_lang: 'auto',
-    censored_words: 'shucks|whiz|whizzer'
+    censored_words: 'shucks|whiz|whizzer',
+    censored_pattern: '\\d{3}-\\d{4}|tech\\w*'
   },
   getURL: url => url
 });
@@ -532,6 +533,9 @@ test("censoring", function() {
   cooked("The link still works. [whiz](http://www.whiz.com)",
          "<p>The link still works. <a href=\"http://www.whiz.com\">&#9632;&#9632;&#9632;&#9632;</a></p>",
          "it won't break links by censoring them.");
+  cooked("Call techapj the computer whiz at 555-555-1234 for free help.",
+         "<p>Call &#9632;&#9632;&#9632;&#9632;&#9632;&#9632;&#9632; the computer &#9632;&#9632;&#9632;&#9632; at 555-&#9632;&#9632;&#9632;&#9632;&#9632;&#9632;&#9632;&#9632; for free help.</p>",
+         "uses both censored words and patterns from site settings");
 });
 
 test("code blocks/spans hoisting", function() {