asciidoctor/asciidoctor-doctest

# frozen_string_literal: true
require 'corefines'
require 'nokogiri'

using Corefines::Object::try

module Asciidoctor
  module DocTest
    ##
    # Module to be included into +Nokogiri::HTML::Document+
    # or +DocumentFragment+ to add {#normalize!} feature.
    #
    # @example
    #   Nokogiri::HTML.parse(str).normalize!
    #   Nokogiri::HTML.fragment(str).normalize!
    module HtmlNormalizer

      ##
      # Normalizes the HTML document or fragment so it can be easily compared
      # with another HTML.
      #
      # What does it actually do?
      #
      # * sorts element attributes by name
      # * sorts inline CSS declarations inside a +style+ attribute by name
      # * removes all blank text nodes (i.e. node that contain just whitespaces)
      # * strips nonsignificant leading and trailing whitespaces around text
      # * strips nonsignificant repeated whitespaces
      #
      # @return [Object] self
      #
      def normalize!
        traverse do |node|
          case node.type

          when Nokogiri::XML::Node::ELEMENT_NODE
            sort_element_attrs! node
            sort_element_style_attr! node

          when Nokogiri::XML::Node::TEXT_NODE
            # Remove text node that contains whitespaces only.
            if node.blank?
              node.remove

            elsif !preformatted_block? node
              strip_redundant_spaces! node
              strip_spaces_around_text! node
            end
          end
        end
        self
      end

      private

      # Sorts attributes of the element +node+ by name.
      def sort_element_attrs!(node)
        node.attributes.sort_by(&:first).each do |name, value|
          node.delete(name)
          node[name] = value
        end
      end

      # Sorts CSS declarations in style attribute of the element +node+ by name.
      def sort_element_style_attr!(node)
        return unless node.has_attribute? 'style'
        decls = node['style'].scan(/([\w-]+):\s*([^;]+);?/).sort_by(&:first)
        node['style'] = decls.map { |name, val| "#{name}: #{val};" }.join(' ')
      end

      # Note: muttable methods like +gsub!+ doesn't work on node content.

      # Strips repeated whitespaces in the text +node+.
      def strip_redundant_spaces!(node)
        node.content = node.content.gsub("\n", ' ').gsub(/(\s)+/, '\1')
      end

      # Strips nonsignificant leading and trailing whitespaces in the text +node+.
      def strip_spaces_around_text!(node)
        node.content = node.content.lstrip if text_block_boundary? node, :left
        node.content = node.content.rstrip if text_block_boundary? node, :right
      end

      ##
      # Returns +true+ if the text +node+ is the first (+:left+), or the last
      # (+:right+) inline element of the nearest block element ancestor or
      # direct sibling of +<br>+ element.
      #
      # @return [Boolean]
      #
      def text_block_boundary?(node, side)
        method = { left: :previous_sibling, right: :next_sibling }[side]

        return true if node.send(method).try(:name) == 'br'
        loop do
          if (sibling = node.send(method))
            return false if sibling.text? || inline_element?(sibling)
          end
          node = node.parent
          return true unless inline_element? node
        end
      end

      HTML_INLINE_ELEMENTS = Nokogiri::HTML::ElementDescription::HTML_INLINE.flatten

      # @return [Boolean] true if the +node+ represents an inline HTML element.
      def inline_element?(node)
        node.element? && HTML_INLINE_ELEMENTS.include?(node.name)
      end

      # @return [Boolean] true if the +node+ is descendant of +<pre>+ node.
      def preformatted_block?(node)
        node.path =~ %r{/pre/}
      end
    end
  end
end

[Nokogiri::HTML::Document, Nokogiri::HTML::DocumentFragment].each do |klass|
  klass.send :include, Asciidoctor::DocTest::HtmlNormalizer
end