lib/rich_text.rb

   1 # frozen_string_literal: true
   2
   3 module RichText
   4   SPAMMY_PHRASES = [
   5     "Business Description:", "Additional Keywords:"
   6   ].freeze
   7
   8   MAX_DESCRIPTION_LENGTH = 500
   9
  10   def self.new(format, text)
  11     case format
  12     when "html" then HTML.new(text || "")
  13     when "markdown" then Markdown.new(text || "")
  14     when "text" then Text.new(text || "")
  15     end
  16   end
  17
  18   class SimpleFormat
  19     include ActionView::Helpers::TextHelper
  20     include ActionView::Helpers::OutputSafetyHelper
  21
  22     def sanitize(text, _options = {})
  23       Sanitize.clean(text, Sanitize::Config::OSM).html_safe
  24     end
  25   end
  26
  27   class Base < String
  28     def spam_score
  29       link_count = 0
  30       link_size = 0
  31
  32       doc = Nokogiri::HTML(to_html)
  33
  34       if doc.content.empty?
  35         link_proportion = 0
  36       else
  37         doc.xpath("//a").each do |link|
  38           link_count += 1
  39           link_size += link.content.length
  40         end
  41
  42         link_proportion = link_size.to_f / doc.content.length
  43       end
  44
  45       spammy_phrases = SPAMMY_PHRASES.count do |phrase|
  46         doc.content.include?(phrase)
  47       end
  48
  49       ([link_proportion - 0.2, 0.0].max * 200) +
  50         (link_count * 40) +
  51         (spammy_phrases * 40)
  52     end
  53
  54     def image
  55       nil
  56     end
  57
  58     def image_alt
  59       nil
  60     end
  61
  62     def description
  63       nil
  64     end
  65
  66     def truncate_html(max_length = nil, img_length = 1000)
  67       html_doc = to_html
  68       return html_doc if max_length.nil?
  69
  70       doc = Nokogiri::HTML::DocumentFragment.parse(html_doc)
  71       keep_or_discards = %w[p h1 h2 h3 h4 h5 h6 pre a table ul ol dl]
  72       accumulated_length = 0
  73       exceeded_node_parent = nil
  74       truncated = false
  75
  76       doc.traverse do |node|
  77         if accumulated_length >= max_length
  78           if node == exceeded_node_parent
  79             exceeded_node_parent = node.parent
  80             node.remove if keep_or_discards.include?(node.name)
  81           else
  82             node.remove
  83           end
  84           next
  85         end
  86
  87         next unless node.children.empty?
  88
  89         if node.text?
  90           accumulated_length += node.text.length
  91         elsif node.name == "img"
  92           accumulated_length += img_length
  93         end
  94
  95         if accumulated_length >= max_length
  96           truncated = true
  97           exceeded_node_parent = node.parent
  98           node.remove
  99         end
 100       end
 101
 102       {
 103         :truncated => truncated,
 104         :html => doc.to_html.html_safe
 105       }
 106     end
 107
 108     protected
 109
 110     def simple_format(text)
 111       SimpleFormat.new.simple_format(text, :dir => "auto")
 112     end
 113
 114     def sanitize(text)
 115       Sanitize.clean(text, Sanitize::Config::OSM).html_safe
 116     end
 117
 118     def linkify(text, mode = :urls)
 119       link_attr = 'rel="nofollow noopener noreferrer" dir="auto"'
 120       Rinku.auto_link(ERB::Util.html_escape(text), mode, link_attr) do |url|
 121         url = shorten_host(url, Settings.linkify_hosts, Settings.linkify_hosts_replacement)
 122         shorten_host(url, Settings.linkify_wiki_hosts, Settings.linkify_wiki_hosts_replacement) do |path|
 123           path.sub(Regexp.new(Settings.linkify_wiki_optional_path_prefix || ""), "")
 124         end
 125       end.html_safe
 126     end
 127
 128     private
 129
 130     def shorten_host(url, hosts, hosts_replacement)
 131       %r{^(https?://([^/]*))(.*)$}.match(url) do |m|
 132         scheme_host, host, path = m.captures
 133         if hosts&.include?(host)
 134           path = yield(path) if block_given?
 135           if hosts_replacement
 136             "#{hosts_replacement}#{path}"
 137           else
 138             "#{scheme_host}#{path}"
 139           end
 140         end || url
 141       end || url
 142     end
 143   end
 144
 145   class HTML < Base
 146     def to_html
 147       linkify(simple_format(self))
 148     end
 149
 150     def to_text
 151       to_s
 152     end
 153   end
 154
 155   class Markdown < Base
 156     def to_html
 157       linkify(sanitize(document.to_html), :all)
 158     end
 159
 160     def to_text
 161       to_s
 162     end
 163
 164     def image
 165       @image_element = first_image_element(document.root) unless defined? @image_element
 166       @image_element.attr["src"] if @image_element
 167     end
 168
 169     def image_alt
 170       @image_element = first_image_element(document.root) unless defined? @image_element
 171       @image_element.attr["alt"] if @image_element
 172     end
 173
 174     def description
 175       return @description if defined? @description
 176
 177       @description = first_truncated_text_content(document.root)
 178     end
 179
 180     private
 181
 182     def document
 183       return @document if @document
 184
 185       @document = Kramdown::Document.new(self)
 186
 187       should_get_dir_auto = lambda do |el|
 188         dir_auto_types = [:p, :header, :codespan, :codeblock, :pre, :ul, :ol, :table, :dl, :math]
 189         return true if dir_auto_types.include?(el.type)
 190         return true if el.type == :a && el.children.length == 1 && el.children[0].type == :text && el.children[0].value == el.attr["href"]
 191
 192         false
 193       end
 194
 195       add_dir = lambda do |element|
 196         element.attr["dir"] ||= "auto" if should_get_dir_auto.call(element)
 197         element.children.each(&add_dir)
 198       end
 199       add_dir.call(@document.root)
 200
 201       @document
 202     end
 203
 204     def first_image_element(element)
 205       return element if image?(element) && element.attr["src"].present?
 206
 207       element.children.find do |child|
 208         nested_image = first_image_element(child)
 209         break nested_image if nested_image
 210       end
 211     end
 212
 213     def first_truncated_text_content(element)
 214       if paragraph?(element)
 215         truncated_text_content(element)
 216       else
 217         element.children.find do |child|
 218           text = first_truncated_text_content(child)
 219           break text unless text.nil?
 220         end
 221       end
 222     end
 223
 224     def truncated_text_content(element)
 225       text = +""
 226
 227       append_text = lambda do |child|
 228         if child.type == :text
 229           text << child.value
 230         else
 231           child.children.each do |c|
 232             append_text.call(c)
 233             break if text.length > MAX_DESCRIPTION_LENGTH
 234           end
 235         end
 236       end
 237       append_text.call(element)
 238
 239       return nil if text.blank?
 240
 241       text.truncate(MAX_DESCRIPTION_LENGTH)
 242     end
 243
 244     def image?(element)
 245       element.type == :img || (element.type == :html_element && element.value == "img")
 246     end
 247
 248     def paragraph?(element)
 249       element.type == :p || (element.type == :html_element && element.value == "p")
 250     end
 251   end
 252
 253   class Text < Base
 254     def to_html
 255       linkify(simple_format(ERB::Util.html_escape(self)))
 256     end
 257
 258     def to_text
 259       to_s
 260     end
 261   end
 262 end