feature: удаление ссылок из текста

https://jira.railsc.ru/browse/SERVICES-597

.gitignore

@@ -7,3 +7,4 @@
 /pkg/
 /spec/reports/
 /tmp/
+/gemfiles

Gemfile

@@ -2,3 +2,4 @@ source 'https://rubygems.org'
 
 # Specify your gem's dependencies in string_tools.gemspec
 gemspec
+

lib/string_tools.rb

@@ -6,6 +6,8 @@
 require 'string_tools/core_ext/string'
 
 module StringTools
+  autoload :HTML, 'string_tools/html'
+
   module CharDet
     # Возвращает true если строка содержит допустимую
     # последовательность байтов для кодировки utf8 и false в обратном случае

lib/string_tools/html.rb

@@ -0,0 +1,60 @@
+# coding: utf-8
+require 'loofah'
+require 'uri'
+
+module StringTools
+  module HTML
+    # Public: Удаляет ссылки на неразрешенные домены
+    #
+    # html    - String содержимое потенциально ненужных ссылок
+    # options - Hash
+    #         :whitelist - Array of String разрешенныe домены
+    #
+    # Examples
+    #   html = '<a href="https://www.yandex.ru">yandex</a>'
+    #
+    #   StringTools::HTML.remove_links(html, whitelist: ['google.com'])
+    #   # => 'yandex'
+    #
+    #   StringTools::HTML.remove_links(html, whitelist: ['yandex.ru'])
+    #   # => '<a href="https://www.yandex.ru">yandex</a>'
+    #
+    #   StringTools::HTML.remove_links(html, whitelist: ['www.yandex.ru'])
+    #   # => '<a href="https://www.yandex.ru">yandex</a>'
+    #
+    #   html = '<a href="https://yandex.ru">yandex</a>'
+    #
+    #   StringTools::HTML.remove_links(html, whitelist: ['www.yandex.ru'])
+    #   # => 'yandex'
+    #
+    # Returns String without links to external resources
+    def self.remove_links(html, options = {})
+      Loofah.fragment(html).scrub!(LinksRemoveScrubber.new(options)).to_s
+    end
+
+    class LinksRemoveScrubber < Loofah::Scrubber
+      def initialize(options)
+        @whitelist = options.fetch(:whitelist)
+      end
+
+      def scrub(node)
+        return unless node.name == 'a'.freeze
+        uri = URI.parse(node['href'.freeze])
+        node.swap(node.children) unless whitelisted? uri.host
+      rescue URI::InvalidURIError => _
+        node.swap(node.children)
+      end
+
+      def whitelisted?(domain)
+        host_parts = domain.split('.'.freeze).reverse!
+        host = host_parts[0] # com, ru ...
+        1.upto(host_parts.length - 1) do |i|
+          subdomain = host_parts[i]
+          host = "#{subdomain}.#{host}"
+          return true if @whitelist.include? host
+        end
+        false
+      end
+    end
+  end
+end

spec/html_spec.rb

@@ -0,0 +1,104 @@
+# coding: utf-8
+require 'spec_helper'
+
+describe StringTools::HTML do
+  describe '.remove_external_links' do
+    context 'whitelist option empty' do
+      subject { StringTools::HTML.remove_links(html, whitelist: []) }
+
+      context 'content without links' do
+        let(:html) { ' <b>hello</b> <script>alert("world")</script> ' }
+
+        it 'should return html as is' do
+          is_expected.to eq html
+        end
+      end
+
+      context 'content with links' do
+        let(:html) do
+          <<-MARKUP
+            <a href="https://google.com"><span>goo</span><span>gle</span></a>
+            <a href="https://yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+
+        it 'should return markup without links' do
+          is_expected.to eq(<<-MARKUP)
+            <span>goo</span><span>gle</span>
+            <span>yan</span><span>dex</span>
+          MARKUP
+        end
+      end
+
+      context 'content with recursive markup' do
+        let(:html) do
+          <<-MARKUP
+            <a href="https://google.com"><a href="https://google.com">goo</a><span>gle</span></a>
+            <a href="https://yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+
+        it 'should return content without links' do
+          is_expected.to eq(<<-MARKUP)
+            goo<span>gle</span>
+            <span>yan</span><span>dex</span>
+          MARKUP
+        end
+      end
+    end
+
+    context 'when whitelist passed' do
+      subject { StringTools::HTML.remove_links(html, whitelist: ['yandex.ru']) }
+
+      context 'domain link match to whitelisted' do
+        let(:html) do
+          <<-MARKUP
+            <a href="https://google.com"><span>goo</span><span>gle</span></a>
+            <a href="https://yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+
+        it 'should keep only whitelisted links' do
+          is_expected.to eq(<<-MARKUP)
+            <span>goo</span><span>gle</span>
+            <a href="https://yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+      end
+
+      context 'link domain is subdomain of whitelisted' do
+        let(:html) do
+          <<-MARKUP
+            <a href="https://google.com"><span>goo</span><span>gle</span></a>
+            <a href="https://www.yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+
+        it 'should keep only whitelisted links' do
+          is_expected.to eq(<<-MARKUP)
+            <span>goo</span><span>gle</span>
+            <a href="https://www.yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+      end
+
+      context 'link domain is parent domain of whitelisted' do
+        subject { StringTools::HTML.remove_links(html, whitelist: ['www.yandex.ru']) }
+
+        let(:html) do
+          <<-MARKUP
+            <a href="https://google.com"><span>goo</span><span>gle</span></a>
+            <a href="https://yandex.ru"><span>yan</span><span>dex</span></a>
+          MARKUP
+        end
+
+        it 'should remove link' do
+          is_expected.to eq(<<-MARKUP)
+            <span>goo</span><span>gle</span>
+            <span>yan</span><span>dex</span>
+          MARKUP
+        end
+      end
+    end
+  end
+end

string_tools.gemspec

@@ -26,13 +26,14 @@ Gem::Specification.new do |spec|
   spec.add_runtime_dependency 'addressable', '~> 2.3.2'
   spec.add_runtime_dependency 'ru_propisju', '~> 2.1.4'
   spec.add_runtime_dependency 'sanitize', '>= 3.1.2'
+  spec.add_runtime_dependency 'loofah', '>= 2.0.0'
 
   spec.add_development_dependency 'bundler', '~> 1.7'
   spec.add_development_dependency 'rake', '~> 10.0'
   spec.add_development_dependency 'rspec', '>= 2.14.0'
   spec.add_development_dependency 'rspec-rails', '>= 2.14.0'
   spec.add_development_dependency 'rspec-given', '~> 3.5'
-  spec.add_development_dependency 'shoulda-matchers'
+  spec.add_development_dependency 'shoulda-matchers', '~> 2.0'
   spec.add_development_dependency 'appraisal', '>= 1.0.2'
   spec.add_development_dependency 'combustion', '>= 0.5.3'
   spec.add_development_dependency 'simplecov', '>= 0.9'