coradoc-html 1.1.18 → 1.1.20

data/lib/coradoc/html/input/converters/text.rb

@@ -1,68 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      module Converters
-        class Text < Base
-          INSTANCE = new
-
-          def to_coradoc(node, state = {})
-            return treat_empty(node, state) if node.text.strip.empty?
-
-            # HTML cleanup is performed in the converter layer
-            cleaned_content = cleanup_html_text(node.text)
-
-            # Return as CoreModel::InlineElement with format_type "text"
-            Coradoc::CoreModel::TextElement.new(
-              content: cleaned_content
-            )
-          end
-
-          private
-
-          def treat_empty(node, state)
-            parent = node.parent.name.to_sym
-            if %i[ol ul].include?(parent) # Otherwise the indentation is broken
-              nil
-            elsif state[:tdsinglepara]
-              nil
-            elsif node.text == ' ' # Regular whitespace text node
-              ' '
-            else
-              nil
-            end
-          end
-
-          # HTML-to-CoreModel text cleanup
-          def cleanup_html_text(text)
-            text = preserve_nbsp(text)
-            text = remove_border_newlines(text)
-            text = remove_inner_newlines(text)
-            escape_links(text)
-          end
-
-          def preserve_nbsp(text)
-            text.gsub("\u00A0", '&nbsp;')
-          end
-
-          def escape_links(text)
-            text.gsub(/<<([^ ][^>]*)>>/, '\\<<\\1>>')
-          end
-
-          def remove_border_newlines(text)
-            text.gsub(/\A\n+/, '').gsub(/\n+\z/, '')
-          end
-
-          def remove_inner_newlines(text)
-            # Convert newlines/tabs to spaces and squeeze multiple spaces
-            # Preserve single leading/trailing space for inline contexts
-            text.tr("\n\t", ' ').squeeze(' ')
-          end
-        end
-
-        register :text, Text::INSTANCE
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/converters/tr.rb

@@ -1,29 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      module Converters
-        class Tr < Base
-          INSTANCE = new
-
-          def to_coradoc(node, state = {})
-            content = treat_children_coradoc(node, state)
-            header = table_header_row?(node)
-            # Use CoreModel::TableRow with cells (not columns)
-            Coradoc::CoreModel::TableRow.new(
-              cells: content,
-              header: header
-            )
-          end
-
-          def table_header_row?(node)
-            node.previous_element.nil?
-          end
-        end
-
-        register :tr, Tr::INSTANCE
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/converters/video.rb

@@ -1,29 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      module Converters
-        class Video < MediaBase
-          INSTANCE = new
-
-          private
-
-          def semantic_type
-            :video
-          end
-
-          def build_attributes(node)
-            base_attributes(node).merge(
-              poster: node['poster'],
-              width: node['width'],
-              height: node['height']
-            ).compact
-          end
-        end
-
-        register :video, Video::INSTANCE
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/converters.rb

@@ -1,107 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      module Converters
-        # Autoload converter classes — they self-register when loaded.
-        # Adding a new converter requires only adding one entry here.
-        CONVERTERS = {
-          Base: 'coradoc/html/input/converters/base',
-          Markup: 'coradoc/html/input/converters/markup',
-          A: 'coradoc/html/input/converters/a',
-          Aside: 'coradoc/html/input/converters/aside',
-          Audio: 'coradoc/html/input/converters/audio',
-          Blockquote: 'coradoc/html/input/converters/blockquote',
-          Br: 'coradoc/html/input/converters/br',
-          Bypass: 'coradoc/html/input/converters/bypass',
-          Code: 'coradoc/html/input/converters/code',
-          Div: 'coradoc/html/input/converters/div',
-          Dl: 'coradoc/html/input/converters/dl',
-          Skip: 'coradoc/html/input/converters/drop',
-          Em: 'coradoc/html/input/converters/em',
-          Figure: 'coradoc/html/input/converters/figure',
-          H: 'coradoc/html/input/converters/h',
-          Head: 'coradoc/html/input/converters/head',
-          Hr: 'coradoc/html/input/converters/hr',
-          Img: 'coradoc/html/input/converters/img',
-          Li: 'coradoc/html/input/converters/li',
-          Mark: 'coradoc/html/input/converters/mark',
-          Math: 'coradoc/html/input/converters/math',
-          MediaBase: 'coradoc/html/input/converters/media_base',
-          Ol: 'coradoc/html/input/converters/ol',
-          P: 'coradoc/html/input/converters/p',
-          PassThrough: 'coradoc/html/input/converters/pass_through',
-          PositionalFormatting: 'coradoc/html/input/converters/positional_formatting',
-          Pre: 'coradoc/html/input/converters/pre',
-          Q: 'coradoc/html/input/converters/q',
-          Strong: 'coradoc/html/input/converters/strong',
-          Sup: 'coradoc/html/input/converters/sup',
-          Sub: 'coradoc/html/input/converters/sub',
-          Table: 'coradoc/html/input/converters/table',
-          Td: 'coradoc/html/input/converters/td',
-          Text: 'coradoc/html/input/converters/text',
-          Tr: 'coradoc/html/input/converters/tr',
-          Video: 'coradoc/html/input/converters/video'
-        }.freeze
-        private_constant :CONVERTERS
-
-        CONVERTERS.each do |name, path|
-          autoload name, path
-        end
-
-        @converters = {}
-        @converters_loaded = false
-
-        def self.register(tag_name, converter)
-          @converters[tag_name.to_sym] = converter
-        end
-
-        def self.unregister(tag_name)
-          @converters.delete(tag_name.to_sym)
-        end
-
-        def self.ensure_converters_loaded
-          return if @converters_loaded
-
-          @converters_loaded = true
-          CONVERTERS.each_key { |name| const_get(name) }
-        end
-
-        def self.lookup(tag_name)
-          ensure_converters_loaded
-          @converters[tag_name.to_sym] || default_converter(tag_name)
-        end
-
-        def self.process_coradoc(node, state)
-          node = node.to_a if node.is_a? Nokogiri::XML::NodeSet
-          return node.map { |i| process_coradoc(i, state) } if node.is_a? Array
-
-          plugins = state[:plugin_instances] || {}
-          process = proc { lookup(node.name).to_coradoc(node, state) }
-          plugins.each do |i|
-            prev_process = process
-            process = proc { i.html_tree_run_hooks(node, state, &prev_process) }
-          end
-          process.call(node, state)
-        end
-
-        def self.default_converter(tag_name)
-          case Html.config.unknown_tags.to_sym
-          when :pass_through
-            PassThrough::INSTANCE
-          when :drop
-            Skip::INSTANCE
-          when :bypass
-            Bypass::INSTANCE
-          when :raise
-            raise Errors::UnknownTagError, "unknown tag: #{tag_name}"
-          else
-            raise Errors::InvalidConfigurationError,
-                  "unknown value #{Html.config.unknown_tags.inspect} for Coradoc::Input::Html.config.unknown_tags"
-          end
-        end
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/errors.rb

@@ -1,22 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      module Errors
-        # Base error class for HTML input errors
-        # Inherits from Coradoc::Error for unified error handling
-        class Error < Coradoc::Error
-        end
-
-        # Raised when an unknown HTML tag is encountered
-        class UnknownTagError < Error
-        end
-
-        # Raised when HTML input configuration is invalid
-        class InvalidConfigurationError < Error
-        end
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/html_converter.rb

@@ -1,98 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      # HTML to CoreModel converter
-      #
-      # This class handles the conversion of HTML documents to CoreModel.
-      # It does NOT handle serialization to any specific output format.
-      # For serialization, use Coradoc.serialize(coremodel, to: :format)
-      #
-      # @example Basic usage - get CoreModel
-      #   coremodel = HtmlConverter.to_core_model(html_string)
-      #
-      # @example Serialize to AsciiDoc
-      #   coremodel = HtmlConverter.to_core_model(html_string)
-      #   adoc_text = Coradoc.serialize(coremodel, to: :asciidoc)
-      #
-      class HtmlConverter
-        # Convert HTML to CoreModel
-        #
-        # @param input [String, Nokogiri::XML::Document, Nokogiri::XML::Node] HTML input
-        # @param options [Hash] Conversion options
-        # @return [Coradoc::CoreModel::Base] CoreModel document
-        def self.to_core_model(input, options = {})
-          Input::Html.config.with(options) do
-            plugin_instances = prepare_plugin_instances(options)
-
-            root = track_time 'Loading input HTML document' do
-              case input
-              when String
-                Nokogiri::HTML(input).root
-              when Nokogiri::XML::Document
-                input.root
-              when Nokogiri::XML::Node
-                input
-              end
-            end
-
-            return nil unless root
-
-            plugin_instances.each do |plugin|
-              plugin.html_tree = root
-              track_time "Preprocessing document with #{plugin.name} plugin" do
-                plugin.preprocess_html_tree
-              end
-              root = plugin.html_tree
-            end
-
-            coremodel = track_time 'Converting input document tree to CoreModel' do
-              Converters.process_coradoc(
-                root,
-                plugin_instances: plugin_instances
-              )
-            end
-
-            coremodel = track_time 'Post-process CoreModel tree' do
-              Postprocessor.process(coremodel)
-            end
-
-            plugin_instances.each do |plugin|
-              plugin.coremodel_tree = coremodel
-              track_time "Postprocessing CoreModel tree with #{plugin.name} plugin" do
-                plugin.postprocess_coremodel_tree
-              end
-              coremodel = plugin.coremodel_tree
-            end
-
-            options[:plugin_instances] = plugin_instances unless options.frozen?
-
-            coremodel
-          end
-        end
-
-        def self.prepare_plugin_instances(options)
-          options[:plugin_instances] || Html.config.plugins.map(&:new)
-        end
-
-        @track_time_indentation = 0
-        def self.track_time(task)
-          if Input::Html.config.track_time
-            warn ('  ' * @track_time_indentation) + "* #{task} is starting..."
-            @track_time_indentation += 1
-            t0 = Time.now
-            ret = yield
-            time_elapsed = Time.now - t0
-            @track_time_indentation -= 1
-            warn ('  ' * @track_time_indentation) +
-                 "* #{task} took #{time_elapsed.round(3)} seconds"
-            ret
-          else
-            yield
-          end
-        end
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/plugin.rb

@@ -1,120 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      class Plugin
-        #### Plugin system general
-
-        # Allow building plugins with a shorthand syntax:
-        # plugin = Coradoc::Html::Input::Plugin.new do
-        #   def name = "Test"
-        # end
-
-        def self.new(&)
-          if self == Plugin
-            Class.new(Plugin, &)
-          else
-            super
-          end
-        end
-
-        def initialize
-          @html_tree_hooks_pre = {}
-          @html_tree_hooks_post = {}
-        end
-
-        def name
-          self.class.name
-        end
-
-        # Default no-op hooks. Plugins override these as needed.
-        def preprocess_html_tree; end
-        def postprocess_coremodel_tree; end
-        def postprocess_output_string; end
-
-        #### HTML Tree functionalities
-
-        attr_accessor :html_tree, :coremodel_tree, :output_string
-
-        def html_tree_change_tag_name_by_css(css, new_name)
-          html_tree.css(css).each do |e|
-            e.name = new_name
-          end
-        end
-
-        def html_tree_change_properties_by_css(css, properties)
-          html_tree.css(css).each do |e|
-            properties.each do |k, v|
-              e[k.to_s] = v
-            end
-          end
-        end
-
-        def html_tree_remove_by_css(css)
-          html_tree.css(css).each(&:remove)
-        end
-
-        def html_tree_replace_with_children_by_css(css)
-          html_tree.css(css).each do |e|
-            e.replace(e.children)
-          end
-        end
-
-        def html_tree_process_to_coremodel(tree, state = {})
-          Coradoc::Html::Input::Converters.process_coradoc(tree, state)
-        end
-
-        # define preprocess_html_tree to process HTML trees
-
-        # Creates a hook to be called instead of converting an element
-        # to a CoreModel node.
-        #
-        # proc |html_node, state|
-        #   coremodel_node
-        # end
-        def html_tree_add_hook_pre(element, &block)
-          @html_tree_hooks_pre[element] = block
-        end
-
-        def html_tree_add_hook_pre_by_css(css, &block)
-          html_tree.css(css).each do |e|
-            html_tree_add_hook_pre(e, &block)
-          end
-        end
-
-        # Creates a hook to be called after converting an element
-        # to a CoreModel node.
-        #
-        # proc |html_node, coremodel_node, state|
-        #   coremodel_node
-        # end
-        def html_tree_add_hook_post(element, &block)
-          @html_tree_hooks_post[element] = block
-        end
-
-        def html_tree_add_hook_post_by_css(css, &block)
-          html_tree.css(css).each do |e|
-            html_tree_add_hook_post(e, &block)
-          end
-        end
-
-        def html_tree_run_hooks(node, state, &)
-          hook_pre = @html_tree_hooks_pre[node]
-          hook_post = @html_tree_hooks_post[node]
-
-          coremodel = hook_pre.call(node, state) if hook_pre
-          coremodel ||= yield node, state
-
-          coremodel = hook_post.call(node, coremodel, state) if hook_post
-
-          coremodel
-        end
-
-        #### CoreModel tree functionalities
-
-        #### Output string functionalities
-      end
-    end
-  end
-end

data/lib/coradoc/html/input/plugins/plateau.rb

@@ -1,214 +0,0 @@
-# frozen_string_literal: true
-
-module Coradoc
-  module Input
-    module Html
-      class Plugin
-        # This plugin enhances documents from the PLATEAU project
-        # to extract more data.
-        #
-        # Usage:
-        # coradoc convert -rcoradoc/input/html/plugins/plateau
-        #   --external-images -u raise --output _out/index.adoc index.html
-        class Plateau < Plugin
-          def name
-            'PLATEAU'
-          end
-
-          def preprocess_html_tree
-            # Remove side menu, so we can generate TOC ourselves
-            html_tree_remove_by_css('.sideMenu')
-
-            # Correct non-semantic classes into semantic HTML tags
-            html_tree_change_tag_name_by_css('.titledata', 'h1')
-            html_tree_change_tag_name_by_css('.subtitledata', 'h2')
-            html_tree_change_tag_name_by_css('.pitemdata', 'h3')
-            html_tree_change_tag_name_by_css('.sitemdata', 'h4')
-            html_tree_change_tag_name_by_css('td[bgcolor="#D0CECE"]', 'th')
-            html_tree_change_tag_name_by_css('td[bgcolor="#d0cece"]', 'th')
-            html_tree_change_tag_name_by_css(
-              '.framedata, .frame_container_box',
-              'aside'
-            )
-            html_tree_change_tag_name_by_css('.frame2data', 'pre')
-            # Assumption that all code snippets in those documents are XML...
-            html_tree_change_properties_by_css(
-              '.frame2data',
-              class: 'brush:xml;'
-            )
-
-            # Remove some CSS ids that are not important to us
-            html_tree_change_properties_by_css('#__nuxt', id: nil)
-            html_tree_change_properties_by_css('#__layout', id: nil)
-            html_tree_change_properties_by_css('#app', id: nil)
-
-            # Handle lists of document 02
-            html_tree_replace_with_children_by_css('.list_num-wrap')
-
-            # Convert table/img caption to become a caption
-            html_tree.css('.imagedata').each do |e|
-              table = e.parent.next&.children&.first
-              if table&.name == 'table'
-                e.name = 'caption'
-                table.prepend_child(e)
-                next
-              end
-
-              img = e.parent.previous&.children&.first
-              next unless img&.name == 'img' && img['src']
-
-              title = e.text.strip
-              img['title'] = title
-              e.remove
-              next
-            end
-
-            # Add hooks for H1, H2, H3, H4
-            html_tree_add_hook_post_by_css(
-              'h1, h2, h3',
-              &method(:handle_headers)
-            )
-            html_tree_add_hook_post_by_css('h4', &method(:handle_headers_h4))
-
-            # Table cells aligned to center
-            html_tree_change_properties_by_css(
-              '.tableTopCenter',
-              align: 'center'
-            )
-
-            # Handle non-semantic lists and indentation
-            html_tree_add_hook_pre_by_css '.text2data' do |node,|
-              text = html_tree_process_to_coremodel(node).strip
-              next '' if text.empty? || text == "\u3000"
-
-              if text.start_with?(/\d+\./)
-                text = text.sub(/\A\d+.\s*/, '')
-                ".. #{text}\n"
-              else
-                text = text.gsub(/^/, '** ')
-                "\n\n//-PT2D\n#{text}\n//-ENDPT2D\n\n"
-              end
-            end
-
-            (3..4).each do |i|
-              html_tree_add_hook_pre_by_css ".text#{i}data" do |node,|
-                text = html_tree_process_to_coremodel(node).strip
-                next '' if text.empty? || text == "\u3000"
-
-                text = text.strip.gsub(/^/, "#{'*' * i} ")
-                "\n\n//-PT#{i}D\n#{text}\n//-ENDPT#{i}D\n\n"
-              end
-            end
-
-            (2..3).each do |i|
-              html_tree_add_hook_pre_by_css ".text#{i}data_point ul" do |node,|
-                text = html_tree_process_to_coremodel(node.children.first.children).strip
-
-                "#{'*' * i} #{text}\n"
-              end
-            end
-
-            (1..20).each do |i|
-              html_tree_add_hook_pre_by_css ".numtextdata_num .list_num#{i}" do |node,|
-                text = html_tree_process_to_coremodel(node).strip
-
-                "[start=#{i}]\n. #{text}\n"
-              end
-            end
-          end
-
-          IM = /[A-Z0-9]{1,3}/
-
-          def handle_headers(node, coradoc, _state)
-            content = coradoc.content.map(&:content).join
-
-            if %w[toc0 toc_0].any? { |i| coradoc.id&.start_with?(i) }
-              # Special content
-              case content.strip
-              when 'はじめに' # Introduction
-                coradoc.style = 'abstract' # The older version document has ".preface"
-                coradoc.level_int = 1
-              when '改定の概要' # Revision overview
-                coradoc.style = 'abstract' # The older version document has ".preface"
-                coradoc.level_int = 1
-              when '参考文献' # Bibliography
-                coradoc.style = 'bibliography'
-                coradoc.level_int = 1
-              when '改訂履歴' # Document history
-                coradoc.style = 'appendix'
-                coradoc.level_int = 1
-              when '0　概要' # Overview
-                coradoc.style = 'abstract' # I'm not sure this is correct
-                coradoc.level_int = 1
-              when '索引' # Index
-                coradoc.style = 'index' # I'm not sure this is correct
-                coradoc.level_int = 1
-              else
-                warn "Unknown section #{content.inspect}"
-              end
-            end
-
-            if node.name == 'h1' && content.start_with?('Annex')
-              coradoc.style = 'appendix'
-              coradoc.content.first.content.sub!(/\AAnnex [A-Z]/, '')
-            end
-
-            # Remove numbers
-            coradoc.content.first.content.sub!(
-              /\A(#{IM}\.)*#{IM}[[:space:]]/o,
-              ''
-            )
-
-            coradoc
-          end
-
-          def handle_headers_h4(_node, coradoc, _state)
-            title = Coradoc.strip_unicode(coradoc.content.first.content)
-            case title
-            when /\A\(\d+\)(.*)/
-              coradoc.level_int = 4
-              coradoc.content.first.content = ::Regexp.last_match(1).strip
-              coradoc
-            when /\A\d+\)(.*)/
-              coradoc.level_int = 5
-              coradoc.content.first.content = ::Regexp.last_match(1).strip
-              coradoc
-            when /\A#{IM}\.#{IM}\.#{IM}\.#{IM}(.*)/o
-              coradoc.level_int = 4
-              coradoc.content.first.content = ::Regexp.last_match(1).strip
-            else
-              if title.empty?
-                # Strip instances of faulty empty paragraphs
-                nil
-              else
-                ["// Unhandled h4 content\n", coradoc]
-              end
-            end
-          end
-
-          def postprocess_output_string
-            str = output_string
-
-            ### Custom indentation handling
-            # If there's a step up, add [none]
-            str = str.gsub(%r{\s+//-ENDPT2D\s+//-PT3D\s+}, "\n[none]\n")
-            str = str.gsub(%r{\s+//-ENDPT2D\s+//-PT4D\s+}, "\n[none]\n")
-            str = str.gsub(%r{\s+//-ENDPT3D\s+//-PT4D\s+}, "\n[none]\n")
-            # Collapse blocks of text[2,3]data
-            str = str.gsub(%r{\s+//-ENDPT[234]D\s+//-PT[234]D\s+}, "\n\n")
-            # In the beginning, add [none]
-            str = str.gsub(%r{\s+//-PT[234]D\s+}, "\n\n[none]\n")
-            # If following with another list, ensure we readd styling
-            str = str.gsub(%r{\s+//-ENDPT[234]D\s+\*}, "\n\n[disc]\n*")
-            # Otherwise, clean up
-            str = str.gsub(%r{\s+//-ENDPT[234]D\s+}, "\n\n")
-
-            self.output_string = str
-          end
-        end
-      end
-    end
-  end
-end
-
-Coradoc::Input::Html.config.plugins << Coradoc::Input::Html::Plugin::Plateau