docbrown/app/services/markdown_processor/parser.rb

module MarkdownProcessor
  class Parser
    BAD_XSS_REGEX = [
      /src=["'](data|&)/i,
      %r{data:text/html[,;][\sa-z0-9]*}i,
    ].freeze

    CODE_BLOCKS_REGEX = /(~{3}|`{3}|`{2}|`)[\s\S]*?\1/

    WORDS_READ_PER_MINUTE = 275.0

    # @param content [String] The user input, mix of markdown and liquid.  This might be an
    #        article's markdown body.
    # @param source [Object] The thing associated with the content.  This might be an article.
    # @param user [User] Who's the one writing the content?
    # @param liquid_tag_options [Hash]
    #
    # @note This is a place to pass in a different policy object.  But like, maybe don't do
    #       that with user input, but perhaps via a data migration script.
    #
    # @see LiquidTagBase for more information regarding the liquid tag options.

    def initialize(content, source: nil, user: nil,
                   liquid_tag_options: {})
      @content = content
      @source = source
      @user = user
      @liquid_tag_options = liquid_tag_options.merge({ source: @source, user: @user })
    end

    # @param prefix_images_options [Hash] params, that need to be passed further to HtmlParser#prefix_all_images
    def finalize(link_attributes: {}, prefix_images_options: { width: 800, synchronous_detail_detection: false })
      options = { hard_wrap: true, filter_html: false, link_attributes: link_attributes }
      renderer = Redcarpet::Render::HTMLRouge.new(options)
      markdown = Redcarpet::Markdown.new(renderer, Constants::Redcarpet::CONFIG)
      catch_xss_attempts(@content)
      code_tag_content = convert_code_tags_to_triple_backticks(@content)
      escaped_content = escape_liquid_tags_in_codeblock(code_tag_content)
      html = markdown.render(escaped_content)
      sanitized_content = ActionController::Base.helpers.sanitize html, { scrubber: RenderedMarkdownScrubber.new }

      begin
        # NOTE: [@rhymes] liquid 5.0.0 does not support ActiveSupport::SafeBuffer,
        # a String substitute, hence we force the conversion before passing it to Liquid::Template.
        # See <https://github.com/Shopify/liquid/issues/1390>
        parsed_liquid = Liquid::Template.parse(sanitized_content.to_str, @liquid_tag_options)

        html = markdown.render(parsed_liquid.render)
      rescue Liquid::SyntaxError => e
        html = e.message
      end

      parse_html(html, prefix_images_options)
    end

    def calculate_reading_time
      word_count = @content.split(/\W+/).count
      (word_count / WORDS_READ_PER_MINUTE).ceil
    end

    def evaluate_markdown(allowed_tags: MarkdownProcessor::AllowedTags::MARKDOWN_PROCESSOR_DEFAULT)
      return if @content.blank?

      renderer = Redcarpet::Render::HTMLRouge.new(hard_wrap: true, filter_html: false)
      markdown = Redcarpet::Markdown.new(renderer, Constants::Redcarpet::CONFIG)
      ActionController::Base.helpers.sanitize(markdown.render(@content),
                                              tags: allowed_tags,
                                              attributes: MarkdownProcessor::AllowedAttributes::MARKDOWN_PROCESSOR)
    end

    def evaluate_limited_markdown(allowed_tags: MarkdownProcessor::AllowedTags::MARKDOWN_PROCESSOR_LIMITED)
      evaluate_markdown(allowed_tags: allowed_tags)
    end

    # rubocop:disable Layout/LineLength
    def evaluate_inline_limited_markdown(allowed_tags: MarkdownProcessor::AllowedTags::MARKDOWN_PROCESSOR_INLINE_LIMITED)
      evaluate_markdown(allowed_tags: allowed_tags)
    end
    # rubocop:enable Layout/LineLength

    def evaluate_listings_markdown(allowed_tags: MarkdownProcessor::AllowedTags::MARKDOWN_PROCESSOR_LISTINGS)
      evaluate_markdown(allowed_tags: allowed_tags)
    end

    def tags_used
      return [] if @content.blank?

      cleaned_parsed = escape_liquid_tags_in_codeblock(@content)
      tags = []
      liquid_tag_options = { source: @source, user: @user }
      Liquid::Template.parse(cleaned_parsed, liquid_tag_options).root.nodelist.each do |node|
        tags << node.class if node.class.superclass.to_s == LiquidTagBase.to_s
      end
      tags.uniq
    rescue Liquid::SyntaxError
      []
    end

    def catch_xss_attempts(markdown)
      markdown_without_code_blocks = markdown.gsub(CODE_BLOCKS_REGEX, "")
      return unless markdown_without_code_blocks.match?(Regexp.union(BAD_XSS_REGEX))

      raise ArgumentError, I18n.t("services.markdown_processor.parser.invalid_markdown_detected")
    end

    def escape_liquid_tags_in_codeblock(content)
      # Escape codeblocks, code spans, and inline code
      content.gsub(/[[:space:]]*~{3}.*?~{3}|[[:space:]]*`{3}.*?`{3}|`{2}.+?`{2}|`{1}.+?`{1}/m) do |codeblock|
        codeblock.gsub!("{% endraw %}", "{----% endraw %----}")
        codeblock.gsub!("{% raw %}", "{----% raw %----}")
        if codeblock.match?(/[[:space:]]*`{3}/)
          "\n{% raw %}\n#{codeblock}\n{% endraw %}\n"
        else
          "{% raw %}#{codeblock}{% endraw %}"
        end
      end
    end

    def convert_code_tags_to_triple_backticks(content)
      # return content if there is not a <code> tag
      return content unless /^<code>$/.match?(content)

      # return content if there is a <pre> and <code> tag
      return content if content.include?("<code>") && content.include?("<pre>")

      # Convert all multiline code tags to triple backticks
      content.gsub(%r{^</?code>$}, "\n```\n")
    end

    private

    def parse_html(html, prefix_images_options)
      return html if html.blank?

      Html::Parser
        .new(html)
        .remove_nested_linebreak_in_list
        .prefix_all_images(**prefix_images_options)
        .wrap_all_images_in_links
        .add_control_class_to_codeblock
        .add_control_panel_to_codeblock
        .add_fullscreen_button_to_panel
        .wrap_all_tables
        .remove_empty_paragraphs
        .escape_colon_emojis_in_codeblock
        .unescape_raw_tag_in_codeblocks
        .wrap_all_figures_with_tags
        .wrap_mentions_with_links
        .html
    end
  end
end