coradoc-html 1.1.18 → 1.1.19

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: be959f8cb4f3c516d0f81d10937dabee6eacae2ddca275df46c8700e9dc800cf
-  data.tar.gz: 13c683ec524a321f1cca99aec03f9659591761ec58cf725cf5aa33ffd4af7705
+  metadata.gz: 011622bc6889a0af8aadea9d1a13e4cd4e322b808f5afe95b5da4a785a4b720f
+  data.tar.gz: ffe601423e5ab805e854e150744983336c49d217858a9f4d5aa660ced64d9f9b
 SHA512:
-  metadata.gz: 02c5c0388481200b0880bb3ab5b8afa042dc7bd32b00ddab298ffe4257186305249e0511b6686f38a57892ccbd99138157ed0b29dc0177cd71a8042a3c085f14
-  data.tar.gz: ce56d6b09705b12f0a8eb333b907d0abef4532e139c22080cc65285157c2c8bc2c693a23fed3eba55141ae9aff1380600b89b3181579c7da4dbf002d1f2b4525
+  metadata.gz: 0225afbdd3e517cb260bad9edc7a469605918937faebcc5a425b202e61ac36a6e66d96304defbca5562aa49421fdb20d436c0d47241ab272a0c19344ae839940
+  data.tar.gz: c2847f907efb900fb4ccef551254e808cc042c7aa5d8af2c0439a493fa560b58e07ed55d643a9c915d4cd6f765b1c7d6cff498e6408fee1e8f242b4833b40ff5

data/lib/coradoc/html/cleaner.rb

@@ -0,0 +1,128 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    class Cleaner
+      INNER_WHITESPACE_REGEX_1 = /\n stem:\[/
+      INNER_WHITESPACE_REGEX_2 = /(stem:\[([^\]]|\\\])*\])\n(?=\S)/
+      NEWLINES_REGEX = /\n{3,}/
+      LEADING_NEWLINE_REGEX = /\A\n+/
+      WHITESPACE_REGEX = /[ \t\r\n]+/
+      TRAILING_WHITESPACE_REGEX = /[ \t\r\n]+\z/
+
+      def tidy(string)
+        return string.transform_values { |i| tidy(i) } if string.is_a? Hash
+
+        result = HtmlConverter.track_time 'Removing inner whitespace' do
+          remove_inner_whitespaces(String.new(string))
+        end
+        result = HtmlConverter.track_time 'Removing newlines' do
+          remove_newlines(result)
+        end
+        result = HtmlConverter.track_time 'Removing leading newlines' do
+          remove_leading_newlines(result)
+        end
+        result = HtmlConverter.track_time 'Cleaning tag borders' do
+          clean_tag_borders(result)
+        end
+        result = HtmlConverter.track_time 'Cleaning punctuation characters' do
+          clean_punctuation_characters(result)
+        end
+        result = remove_block_leading_newlines(result)
+        result = remove_section_attribute_newlines(result)
+      end
+
+      def remove_block_leading_newlines(string)
+        string.gsub("]\n****\n\n", "]\n****\n")
+      end
+
+      def remove_section_attribute_newlines(string)
+        string.gsub("]\n\n==", "]\n==")
+      end
+
+      def remove_newlines(string)
+        string.gsub(NEWLINES_REGEX, "\n\n")
+      end
+
+      def remove_leading_newlines(string)
+        string.gsub(LEADING_NEWLINE_REGEX, '')
+      end
+
+      def remove_inner_whitespaces(string)
+        unless string.nil?
+          string.gsub!("\n stem:[", "\nstem:[")
+          string.gsub!(INNER_WHITESPACE_REGEX_1, '\\1 ')
+          string.gsub!(INNER_WHITESPACE_REGEX_2, '\\1')
+        end
+        result = +''
+        string.each_line do |line|
+          result << preserve_border_whitespaces(line) do
+            line.gsub(/\A[ \t\r\n]+/, '').gsub(/[ \t\r\n]+\z/, '').gsub(/[ \t]{2,}/, ' ')
+          end
+        end
+        result
+      end
+
+      def clean_tag_borders(string)
+        result = string.gsub(/\s?~{2,}.*?~{2,}\s?/) do |match|
+          preserve_border_whitespaces(
+            match,
+            default_border: Html.input_config.tag_border
+          ) do
+            match.strip.sub('~~ ', '~~').sub(' ~~', '~~')
+          end
+        end
+
+        result.gsub(/\s?\[.*?\]\s?/) do |match|
+          preserve_border_whitespaces(match) do
+            match.strip.sub('[ ', '[').sub(' ]', ']')
+          end
+        end
+      end
+
+      def clean_punctuation_characters(string)
+        string.gsub(/(\*\*|~~|__)\s([.!?'"])/, '\\1\\2')
+      end
+
+      def preprocess_word_html(string)
+        clean_headings(scrub_whitespace(string.dup))
+      end
+
+      def scrub_whitespace(string)
+        string.gsub!(/&nbsp;|&#xA0;| /i, '&#xA0;')
+        string = Coradoc.strip_unicode(string)
+        string.gsub!(/( +)$/, ' ')
+        string.gsub!("\n\n\n\n", "\n\n")
+        string
+      end
+
+      def clean_headings(string)
+        string.gsub!(%r{<h([1-9])[^>]*></h\1>}, ' ')
+        string.gsub!(
+          %r{<h([1-9])[^>]* style="vertical-align: super;[^>]*>(.+?)</h\1>},
+          '<sup>\\2</sup>'
+        )
+        string
+      end
+
+      private
+
+      def preserve_border_whitespaces(string, options = {})
+        return string if /\A\s*\Z/.match?(string)
+
+        default_border = options.fetch(:default_border, '')
+        default_border = '' if /[\[(\])]/.match?(string)
+        string_start   = present_or_default(string[/\A\s*/], default_border)
+        string_end     = present_or_default(string[/\s*\Z/], default_border)
+        result         = yield
+        string_start + result + string_end
+      end
+
+      def present_or_default(string, default)
+        return default if string.nil? || string.empty?
+
+        string
+      end
+    end
+  end
+end

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

@@ -0,0 +1,77 @@
+# frozen_string_literal: true
+
+require 'coradoc'
+
+module Coradoc
+  module Html
+    module Converters
+      class A < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          # Use treat_children_coradoc to get CoreModel elements
+          content = treat_children_coradoc(node, state)
+
+          href  = node['href']
+          title = extract_title(node)
+          id = node['id'] || node['name']
+
+          id = id&.gsub(/\s/, '')&.gsub(/__+/, '_')
+          id = nil if id&.empty?
+
+          return nil if /^_Toc\d+$|^_GoBack$/.match?(id)
+
+          # For inline anchors - return CoreModel InlineElement with format_type "anchor"
+          if id
+            return Coradoc::CoreModel::InlineElement.new(
+              format_type: 'anchor',
+              target: id
+            )
+          end
+
+          # For cross-references
+          if href.to_s.start_with?('#')
+            ref_id = href.sub(/^#/, '').gsub(/\s/, '').gsub(/__+/, '_')
+            content_str = extract_text_from_content(content)
+            return Coradoc::CoreModel::CrossReferenceElement.new(
+              target: ref_id,
+              content: content_str.strip.empty? ? nil : content_str.strip
+            )
+          end
+
+          return nil if href.to_s.empty?
+
+          # For links
+          ambigous_characters = /[\w.?&#=%;\[\u{ff}-\u{10ffff}]/
+          right_constrain = textnode_after_start_with?(node, ambigous_characters)
+
+          content_str = extract_text_from_content(content)
+
+          out = []
+          # Add leading space if needed
+          if textnode_before_end_with?(node, ambigous_characters)
+            out << Coradoc::CoreModel::TextElement.new(
+              content: ' '
+            )
+          end
+
+          # Create link element
+          link = Coradoc::CoreModel::LinkElement.new(
+            target: href,
+            content: content_str.strip,
+            metadata: {
+              title: (title.strip unless title.to_s.strip.empty?),
+              right_constrain: right_constrain
+            }.compact
+          )
+          out << link
+
+          # Return single element or array
+          out.length == 1 ? out.first : out
+        end
+      end
+
+      register :a, A::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,20 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Aside < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          content = treat_children_coradoc(node, state)
+          Coradoc::CoreModel::SidebarBlock.new(
+            children: content
+          )
+        end
+      end
+
+      register :aside, Aside::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Audio < MediaBase
+        INSTANCE = new
+
+        private
+
+        def semantic_type
+          :audio
+        end
+      end
+
+      register :audio, Audio::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,98 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Base
+        def treat_children_coradoc(node, state)
+          results = node.children.map do |child|
+            treat_coradoc(child, state)
+          end.flatten
+
+          results.reject do |x|
+            x.nil? || (x.is_a?(String) && x.strip.empty?)
+          end
+        end
+
+        def treat_coradoc(node, state)
+          Converters.process_coradoc(node, state)
+        end
+
+        def extract_title(node)
+          node['title'].to_s
+        end
+
+        def node_has_ancestor?(node, name)
+          case name
+          when String
+            node.ancestors(name).any?
+          when Array
+            name.any? { |n| node.ancestors(n).any? }
+          end
+        end
+
+        def textnode_before_end_with?(node, str)
+          return false unless [String, Regexp].include?(str.class)
+          return false if str.is_a?(String) && str.empty?
+
+          str = /#{Regexp.escape(str)}/ if str.is_a?(String)
+          str = /(?:#{str})\z/
+
+          node2 = node.at_xpath('preceding-sibling::node()[1]')
+          node2.is_a?(Nokogiri::XML::Text) && node2.text.match?(str)
+        end
+
+        def textnode_after_start_with?(node, str)
+          return false unless [String, Regexp].include?(str.class)
+          return false if str.is_a?(String) && str.empty?
+
+          str = /#{Regexp.escape(str)}/ if str.is_a?(String)
+          str = /\A(?:#{str})/
+
+          node2 = node.at_xpath('following-sibling::node()[1]')
+          node2.is_a?(Nokogiri::XML::Text) && node2.text.match?(str)
+        end
+
+        def extract_leading_trailing_whitespace(node)
+          node.text =~ /^(\s+)/
+          leading_whitespace = ::Regexp.last_match(1)
+          unless leading_whitespace.nil?
+            first_text = node.at_xpath('./text()[1]')
+            first_text&.replace(first_text.text.lstrip)
+            leading_whitespace = ' '
+          end
+          node.text =~ /(\s+)$/
+          trailing_whitespace = ::Regexp.last_match(1)
+          unless trailing_whitespace.nil?
+            last_text = node.at_xpath('./text()[last()]')
+            last_text&.replace(last_text.text.rstrip)
+            trailing_whitespace = ' '
+          end
+          [leading_whitespace, trailing_whitespace]
+        end
+
+        def unconstrained_before?(node)
+          before = node.at_xpath('preceding::node()[1]')
+
+          before &&
+            !before.text.strip.empty? &&
+            before.text[-1]&.match?(/\w/)
+        end
+
+        def unconstrained_after?(node)
+          after = node.at_xpath('following::node()[1]')
+
+          after && !after.text.strip.empty? &&
+            after.text[0]&.match?(/\w|,|;|"|\.\?!/)
+        end
+
+        # Extract plain text from a mixed content array. Delegates to
+        # CoreModel::InlineContent.text_of — single source of truth for
+        # nil/Array/InlineElement/StructuralElement handling.
+        def extract_text_from_content(content)
+          Coradoc::CoreModel::InlineContent.text_of(content)
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Blockquote < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          id = node['id']
+          cite = node['cite']
+          content = treat_children_coradoc(node, state)
+
+          Coradoc::CoreModel::QuoteBlock.new(
+            children: content,
+            id: id,
+            attribution: cite
+          )
+        end
+      end
+
+      register :blockquote, Blockquote::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Br < Base
+        INSTANCE = new
+
+        def to_coradoc(_node, _state = {})
+          Coradoc::CoreModel::LineBreakElement.new
+        end
+      end
+
+      register :br, Br::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,82 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Bypass < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          treat_children_coradoc(node, state)
+        end
+      end
+
+      register :document, Bypass::INSTANCE
+      register :html,     Bypass::INSTANCE
+      register :body,     Bypass::INSTANCE
+      register :span,     Bypass::INSTANCE
+      register :thead,    Bypass::INSTANCE
+      register :tbody,    Bypass::INSTANCE
+      register :tfoot,    Bypass::INSTANCE
+      register :abbr,     Bypass::INSTANCE
+      register :acronym,  Bypass::INSTANCE
+      register :address,  Bypass::INSTANCE
+      register :applet,   Bypass::INSTANCE
+      register :map,      Bypass::INSTANCE
+      register :area,     Bypass::INSTANCE
+      register :bdi,      Bypass::INSTANCE
+      register :bdo,      Bypass::INSTANCE
+      register :big,      Bypass::INSTANCE
+      register :button,   Bypass::INSTANCE
+      register :canvas,   Bypass::INSTANCE
+      register :data,     Bypass::INSTANCE
+      register :datalist, Bypass::INSTANCE
+      register :del,      Bypass::INSTANCE
+      register :ins,      Bypass::INSTANCE
+      register :dfn,      Bypass::INSTANCE
+      register :dialog,   Bypass::INSTANCE
+      register :embed,    Bypass::INSTANCE
+      register :fieldset, Bypass::INSTANCE
+      register :font,     Bypass::INSTANCE
+      register :footer,   Bypass::INSTANCE
+      register :form,     Bypass::INSTANCE
+      register :frame,    Bypass::INSTANCE
+      register :frameset, Bypass::INSTANCE
+      register :header,   Bypass::INSTANCE
+      register :iframe,   Bypass::INSTANCE
+      register :input,    Bypass::INSTANCE
+      register :label,    Bypass::INSTANCE
+      register :legend,   Bypass::INSTANCE
+      register :main,     Bypass::INSTANCE
+      register :menu,     Bypass::INSTANCE
+      register :menulist, Bypass::INSTANCE
+      register :meter,    Bypass::INSTANCE
+      register :nav,      Bypass::INSTANCE
+      register :noframes, Bypass::INSTANCE
+      register :noscript, Bypass::INSTANCE
+      register :object,   Bypass::INSTANCE
+      register :optgroup, Bypass::INSTANCE
+      register :option,   Bypass::INSTANCE
+      register :output,   Bypass::INSTANCE
+      register :param,    Bypass::INSTANCE
+      register :picture,  Bypass::INSTANCE
+      register :progress, Bypass::INSTANCE
+      register :ruby,     Bypass::INSTANCE
+      register :rt,       Bypass::INSTANCE
+      register :rp,       Bypass::INSTANCE
+      register :s,        Bypass::INSTANCE
+      register :select,   Bypass::INSTANCE
+      register :small,    Bypass::INSTANCE
+      register :strike,   Bypass::INSTANCE
+      register :details,  Bypass::INSTANCE
+      register :section,  Bypass::INSTANCE
+      register :summary,  Bypass::INSTANCE
+      register :svg,      Bypass::INSTANCE
+      register :template, Bypass::INSTANCE
+      register :textarea, Bypass::INSTANCE
+      register :track,    Bypass::INSTANCE
+      register :u,        Bypass::INSTANCE
+      register :wbr,      Bypass::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Code < Markup
+        INSTANCE = new
+
+        def coradoc_format_type
+          'monospace'
+        end
+
+        def markup_ancestor_tag_names
+          %w[code tt kbd samp var]
+        end
+      end
+
+      register :code, Code::INSTANCE
+      register :tt,   Code::INSTANCE
+      register :kbd,  Code::INSTANCE
+      register :samp, Code::INSTANCE
+      register :var,  Code::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,23 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Div < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          id = node['id']
+          contents = treat_children_coradoc(node, state)
+
+          Coradoc::CoreModel::OpenBlock.new(
+            children: contents,
+            id: id
+          )
+        end
+      end
+
+      register :div, Div::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,82 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Dl < Base
+        INSTANCE = new
+
+        def to_coradoc(node, state = {})
+          items = process_dl(node, state)
+
+          # Convert items to CoreModel::ListItem objects
+          # For definition lists, term goes in content, definition goes in children
+          list_items = items.map do |item|
+            term_text = extract_text_from_content(item[:name])
+            Coradoc::CoreModel::ListItem.new(
+              content: term_text,
+              children: item[:value]
+            )
+          end
+
+          # Use CoreModel::ListBlock with marker_type "definition"
+          Coradoc::CoreModel::ListBlock.new(
+            marker_type: 'definition',
+            items: list_items
+          )
+        end
+
+        def process_dl(node, state = {})
+          groups = []
+          current = { name: [], value: [] }
+
+          seen_dd = false
+          child = node.at_xpath('*[1]')
+          grandchild = nil
+          until child.nil?
+            if child.name == 'div'
+              grandchild = child.at_xpath('*[1]')
+              until grandchild.nil?
+                groups, current, seen_dd = process_dt_or_dd(
+                  groups,
+                  current,
+                  seen_dd,
+                  grandchild,
+                  state
+                )
+                grandchild = grandchild.at_xpath('following-sibling::*[1]')
+              end
+            elsif %w[dt dd].include?(child.name)
+              groups, current, seen_dd = process_dt_or_dd(
+                groups,
+                current,
+                seen_dd,
+                child,
+                state
+              )
+            end
+            child = child.at_xpath('following-sibling::*[1]')
+            groups << current if current[:name].any? && current[:value].any?
+          end
+          groups
+        end
+
+        def process_dt_or_dd(groups, current, seen_dd, subnode, state = {})
+          if subnode.name == 'dt'
+            if seen_dd
+              current = { name: [], value: [] }
+              seen_dd = false
+            end
+            current[:name] += treat_children_coradoc(subnode, state)
+          elsif subnode.name == 'dd'
+            current[:value] += treat_children_coradoc(subnode, state)
+            seen_dd = true
+          end
+          [groups, current, seen_dd]
+        end
+      end
+
+      register :dl, Dl::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Skip < Base
+        INSTANCE = new
+
+        def to_coradoc(_node, _state = {})
+          ''
+        end
+      end
+
+      register :caption, Skip::INSTANCE
+      register :figcaption, Skip::INSTANCE
+      register :title,     Skip::INSTANCE
+      register :link,      Skip::INSTANCE
+      register :style,     Skip::INSTANCE
+      register :meta,      Skip::INSTANCE
+      register :script,    Skip::INSTANCE
+      register :comment,   Skip::INSTANCE
+      register :colgroup,  Skip::INSTANCE
+      register :col,       Skip::INSTANCE
+    end
+  end
+end

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

@@ -0,0 +1,23 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Html
+    module Converters
+      class Em < Markup
+        INSTANCE = new
+
+        def coradoc_format_type
+          'italic'
+        end
+
+        def markup_ancestor_tag_names
+          %w[em i cite]
+        end
+      end
+
+      register :em,   Em::INSTANCE
+      register :i,    Em::INSTANCE
+      register :cite, Em::INSTANCE
+    end
+  end
+end