docbrown/app/services/feeds/assemble_article_markdown.rb

module Feeds
  class AssembleArticleMarkdown
    def self.call(item, user, feed, feed_source_url)
      new(item, user, feed, feed_source_url).call
    end

    def initialize(item, user, feed, feed_source_url)
      @item = item
      @title = item.title.strip
      @categories = item.categories || []
      @user = user
      @feed = feed
      @feed_source_url = feed_source_url
    end

    def call
      body = <<~HEREDOC
        ---
        title: #{processed_title}
        published: false
        date: #{@item.published}
        tags: #{get_tags}
        canonical_url: #{@user.setting.feed_mark_canonical ? @feed_source_url : ''}
        ---

        #{assemble_body_markdown}
      HEREDOC

      body.strip
    end

    private

    def processed_title
      @title.truncate(128, omission: "...", separator: " ")
    end

    def get_tags
      @categories.first(4).map do |tag|
        tag.delete(" ").gsub(/[^[:alnum:]]/i, "")[0..19]
      end.join(",")
    end

    def assemble_body_markdown
      cleaned_content = Feeds::CleanHtml.call(get_content)
      cleaned_content = thorough_parsing(cleaned_content, @feed.url)

      content = ReverseMarkdown
        .convert(cleaned_content, github_flavored: true)
        .gsub("```\n\n```", "")
        .gsub(/&nbsp;|\u00A0/, " ")

      content.gsub!(/{%\syoutube\s(.{11,18})\s%}/) do |tag|
        tag.gsub("\\_", "_")
      end

      content
    end

    def get_content
      @item.content || @item.summary || @item.description
    end

    def thorough_parsing(content, feed_url)
      html_doc = Nokogiri::HTML(content)

      find_and_replace_possible_links!(html_doc) if @user.setting.feed_referential_link
      find_and_replace_picture_tags_with_img!(html_doc)

      if feed_url&.include?("medium.com")
        parse_and_translate_gist_iframe!(html_doc)
        parse_and_translate_youtube_iframe!(html_doc)
        parse_and_translate_tweet!(html_doc)
        parse_liquid_variable!(html_doc)
      elsif feed_url
        clean_relative_path!(html_doc, feed_url)
      end

      html_doc.to_html
    end

    def parse_and_translate_gist_iframe!(html_doc)
      html_doc.css("iframe").each do |iframe|
        a_tag = iframe.css("a")
        next if a_tag.empty?

        possible_link = a_tag[0].inner_html
        next unless %r{medium\.com/media/.+/href}.match?(possible_link)

        real_link = HTTParty.head(possible_link).request.last_uri.to_s
        return nil unless real_link.include?("gist.github.com")

        iframe.name = "p"
        iframe.keys.each { |attr| iframe.remove_attribute(attr) } # rubocop:disable Style/HashEachMethods
        iframe.inner_html = "{% gist #{real_link} %}"
      end
      html_doc
    end

    def parse_and_translate_tweet!(html_doc)
      html_doc.search("style").remove
      html_doc.search("script").remove
      html_doc.css("blockquote").each do |bq|
        bq_with_p = bq.css("p")

        next if bq_with_p.empty?
        next unless (tweet_link = bq_with_p.css("a[href*='twitter.com']"))

        bq.name = "p"
        tweet_url = tweet_link.attribute("href").value
        tweet_id = tweet_url.split("/status/").last
        bq.inner_html = "{% tweet #{tweet_id} %}"
      end
    end

    def parse_liquid_variable!(html_doc)
      # Medium articles does not wrap {{ }} content in liquid tag.
      # This will wrap do so for content that isn't in pre and code tag
      html_doc.css("//body :not(pre):not(code)").each do |node|
        node.inner_html = node.inner_html.gsub(/{{.*?}}/) { |liquid| "`#{liquid}`" }
      end
    end

    def parse_and_translate_youtube_iframe!(html_doc)
      html_doc.css("iframe").each do |iframe|
        next unless /youtube\.com/.match?(iframe.attributes["src"].value)

        iframe.name = "p"
        youtube_id = iframe.attributes["src"].value.scan(/embed%2F(.{4,11})/).flatten.first
        iframe.keys.each { |attr| iframe.remove_attribute(attr) } # rubocop:disable Style/HashEachMethods
        iframe.inner_html = "{% youtube #{youtube_id} %}"
      end
    end

    def clean_relative_path!(html_doc, url)
      html_doc.css("img").each do |img_tag|
        path = (img_tag.attributes["src"] || img_tag.attributes["data-src"])&.value
        next unless path

        # Only update source if the path is not already an URL
        unless path.match?(/\A#{URI::DEFAULT_PARSER.make_regexp}\z/)
          resource = path.start_with?("/") ? url : @feed_source_url
          img_tag.attributes["src"].value = URI.join(resource, path).to_s
        end
      end
    end

    def find_and_replace_possible_links!(html_doc)
      html_doc.css("a").each do |a_tag|
        link = a_tag.attributes["href"]&.value
        next unless link

        found_article = Article.find_by(feed_source_url: link)&.decorate
        a_tag.attributes["href"].value = found_article.url if found_article
      end
    end

    # <picture> tags are not automatically converted to Markdown, they live on as verbatim HTML.
    # Since they are also not supported by the editor, we need to replace them with their inner <img> tag.
    # We'll rely on Cloudinary to provide responsive images to the client
    def find_and_replace_picture_tags_with_img!(html_doc)
      picture_tags = html_doc.css("picture")
      return unless picture_tags

      picture_tags.each do |picture_tag|
        img_tag = picture_tag.at_css("img")
        next unless img_tag

        picture_tag.add_next_sibling(img_tag)
        picture_tag.remove
      end
    end
  end
end