coradoc-docx 0.1.0

data/lib/coradoc/docx/transform/style_resolver.rb

@@ -0,0 +1,249 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Docx
+    module Transform
+      # Resolves paragraph and run styles to semantic roles.
+      #
+      # OOXML paragraphs don't have explicit element types. Instead, their
+      # meaning is determined by style references (e.g., "Heading1" → section)
+      # or by formatting properties (e.g., numPr → list item).
+      #
+      # StyleResolver centralizes this detection so HeadingRule, ListItemRule,
+      # and ParagraphRule don't duplicate the logic.
+      #
+      # The style map is built from the Uniword StylesConfiguration by walking
+      # all style definitions and their basedOn chains.
+      class StyleResolver
+        HEADING_PATTERN = /^(heading|heading|h)\s*(\d+)$/i
+        QUOTE_PATTERN = /\bquote\b/i
+        CODE_PATTERN = /\b(code|source|listing)\b/i
+        LITERAL_PATTERN = /\bliteral\b/i
+        EXAMPLE_PATTERN = /\bexample\b/i
+
+        # @param styles_configuration [Object, nil] Uniword styles configuration
+        def initialize(styles_configuration)
+          @config = styles_configuration
+          @style_map = build_style_map(styles_configuration)
+        end
+
+        # Determine the semantic role of a paragraph
+        #
+        # @param paragraph [Uniword::Wordprocessingml::Paragraph]
+        # @return [Symbol] :heading, :list_item, :quote, :source, :literal,
+        #   :example, or :paragraph
+        def semantic_role(paragraph)
+          return :heading if heading?(paragraph)
+          return :list_item if list_item?(paragraph)
+
+          style_role = role_from_style(paragraph)
+          return style_role if style_role
+
+          :paragraph
+        end
+
+        # Check if paragraph is a heading
+        # @param paragraph [Uniword::Wordprocessingml::Paragraph]
+        # @return [Boolean]
+        def heading?(paragraph)
+          return false unless paragraph.properties
+
+          style_name = resolve_style_name(paragraph)
+          return true if style_name && HEADING_PATTERN.match?(style_name)
+
+          ol = paragraph.properties.outline_level
+          if ol
+            ol_level = ol.is_a?(Uniword::Wordprocessingml::OutlineLevel) ? ol.value.to_i : ol.to_i
+            return true if ol_level.positive?
+          end
+
+          style = find_style_for_paragraph(paragraph)
+          if style&.outline_level
+            ol_val = style.outline_level
+            ol_val = ol_val.is_a?(Uniword::Wordprocessingml::OutlineLevel) ? ol_val.value.to_i : ol_val.to_i
+            return true if ol_val.positive?
+          end
+
+          false
+        end
+
+        # Get heading level (1-6) or nil
+        # @param paragraph [Uniword::Wordprocessingml::Paragraph]
+        # @return [Integer, nil]
+        def heading_level(paragraph)
+          style_name = resolve_style_name(paragraph)
+          if style_name
+            match = HEADING_PATTERN.match(style_name)
+            return match[2].to_i if match
+          end
+
+          # Check outline_level on paragraph properties
+          ol = paragraph.properties&.outline_level
+          if ol
+            level = ol.is_a?(Uniword::Wordprocessingml::OutlineLevel) ? ol.value.to_i : ol.to_i
+            return level if level.positive?
+          end
+
+          nil
+        end
+
+        # Check if paragraph is a list item
+        # @param paragraph [Uniword::Wordprocessingml::Paragraph]
+        # @return [Boolean]
+        def list_item?(paragraph)
+          return false unless paragraph.properties
+
+          num_id = paragraph.properties.num_id
+          num_id.to_i.positive?
+        end
+
+        # Check if paragraph has a specific role based on style name
+        # @param paragraph [Uniword::Wordprocessingml::Paragraph]
+        # @return [Symbol, nil]
+        def role_from_style(paragraph)
+          style_name = resolve_style_name(paragraph)
+          return nil unless style_name
+
+          case style_name
+          when QUOTE_PATTERN then :quote
+          when CODE_PATTERN then :source
+          when LITERAL_PATTERN then :literal
+          when EXAMPLE_PATTERN then :example
+          end
+        end
+
+        # Detect semantic role of a run based on its rStyle
+        # @param run [Uniword::Wordprocessingml::Run]
+        # @return [Symbol, nil]
+        def run_semantic_role(run)
+          return nil unless run.properties
+          return nil unless run.properties.style
+
+          style_name = resolve_run_style_name(run)
+          return nil unless style_name
+
+          case style_name
+          when /\b(code|verbatim|teletype|keyboard)\b/i then :monospace
+          when /\bstrong\b/i then :bold
+          when /\b(emphasis|em)\b/i then :italic
+          when /\bcitation\b/i then :italic
+          end
+        end
+
+        private
+
+        def resolve_style_name(paragraph)
+          style_ref = paragraph.properties&.style
+          return nil unless style_ref
+
+          value = style_ref.is_a?(Uniword::Wordprocessingml::PStyle) ? style_ref.val : style_ref.to_s
+          return nil unless value
+
+          mapped = @style_map[value]
+          return mapped if mapped
+
+          value
+        end
+
+        def resolve_run_style_name(run)
+          style_ref = run.properties.style
+          return nil unless style_ref
+
+          value = style_ref.is_a?(Uniword::Wordprocessingml::PStyle) ? style_ref.val : style_ref.to_s
+          return nil unless value
+
+          mapped = @style_map[value]
+          mapped || value
+        end
+
+        def find_style_for_paragraph(paragraph)
+          return nil unless @config
+
+          style_id = style_id_from_paragraph(paragraph)
+          return nil unless style_id
+
+          return unless @config.is_a?(Uniword::Wordprocessingml::StylesConfiguration)
+
+          @config.style_by_id(style_id)
+        end
+
+        def style_id_from_paragraph(paragraph)
+          style_ref = paragraph.properties&.style
+          return nil unless style_ref
+
+          style_ref.is_a?(Uniword::Wordprocessingml::PStyle) ? style_ref.val : style_ref.to_s
+        end
+
+        def build_style_map(config)
+          return {} unless config
+          return {} unless config.is_a?(Uniword::Wordprocessingml::StylesConfiguration)
+
+          map = {}
+          config.styles.each do |style|
+            id = style.styleId
+            name = extract_style_name(style)
+            next unless id && name
+
+            map[id] = name
+
+            if heading_by_based_on?(config, style)
+              level = heading_level_from_chain(config, style)
+              map[id] = "Heading#{level}"
+            end
+          end
+
+          map
+        end
+
+        def extract_style_name(style)
+          sn = style.style_name
+          return sn if sn
+
+          name = style.name
+          return nil unless name
+
+          name.is_a?(Uniword::Wordprocessingml::StyleName) ? name.val.to_s : name.to_s
+        end
+
+        def heading_by_based_on?(config, style)
+          based_on = style.based_on
+          return false unless based_on
+
+          visited = Set.new
+          current = style
+          while current && !visited.include?(current.styleId)
+            visited << current.styleId
+            parent_id = current.based_on
+            return true if parent_id && HEADING_PATTERN.match?(parent_id)
+
+            break unless parent_id
+
+            current = config.styles.find { |s| s.styleId == parent_id }
+          end
+
+          false
+        end
+
+        def heading_level_from_chain(config, style)
+          visited = Set.new
+          current = style
+          while current && !visited.include?(current.styleId)
+            visited << current.styleId
+            name = extract_style_name(current)
+            if name
+              match = HEADING_PATTERN.match(name)
+              return match[2].to_i if match
+            end
+
+            parent_id = current.based_on
+            break unless parent_id
+
+            current = config.styles.find { |s| s.styleId == parent_id }
+          end
+
+          1
+        end
+      end
+    end
+  end
+end

data/lib/coradoc/docx/transform/to_core_model.rb

@@ -0,0 +1,340 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Docx
+    module Transform
+      # Orchestrator for OOXML → CoreModel transformation.
+      #
+      # Walks a Uniword::Wordprocessingml::DocumentRoot tree and dispatches
+      # to registered transform rules. Handles:
+      #
+      # - Style-based heading detection (via StyleResolver)
+      # - List grouping (consecutive numPr paragraphs → single ListBlock)
+      # - Footnote content collection
+      # - Image reference tracking
+      # - Bookmark ID propagation
+      #
+      # Dispatch strategy:
+      # - HeadingRule and ListItemRule are dispatched directly by the
+      #   orchestrator (they need context for style resolution).
+      # - All other element types are dispatched via RuleRegistry.
+      #
+      # @example Transform a DOCX document
+      #   doc = Uniword::DocumentFactory.from_file("input.docx")
+      #   core = ToCoreModel.transform(doc)
+      #   # => Coradoc::CoreModel::StructuralElement
+      class ToCoreModel
+        class << self
+          def transform(document)
+            new.transform(document)
+          end
+        end
+
+        def transform(document)
+          registry = build_registry
+
+          context = Context.new(
+            styles_configuration: document.styles_configuration,
+            numbering_configuration: document.numbering_configuration,
+            footnotes: collect_footnotes(document),
+            registry: registry
+          )
+
+          @heading_rule = Rules::HeadingRule.new
+          @list_item_rule = Rules::ListItemRule.new
+
+          body = document.body
+          doc_title = extract_document_title(document, context)
+          children = transform_elements(body, context)
+
+          # If the first child is an H1 matching the doc title, skip the
+          # duplicate — the document title already captures it
+          if doc_title && children.first.is_a?(Coradoc::CoreModel::StructuralElement) &&
+             children.first.section? &&
+             children.first.title == doc_title &&
+             children.first.level == 1
+            children.shift
+          end
+
+          doc = Coradoc::CoreModel::StructuralElement.new(
+            element_type: 'document',
+            title: doc_title,
+            children: children
+          )
+
+          # Extract semantic content from headers/footers
+          extract_header_footer_metadata(document, doc)
+
+          doc
+        end
+
+        private
+
+        # Walk body elements with list grouping look-ahead
+        def transform_elements(body, context)
+          return [] unless body
+
+          elements = body_ordered_elements(body)
+
+          result = []
+          i = 0
+
+          while i < elements.length
+            element = elements[i]
+
+            transformed = dispatch_element(element, i, elements, context)
+
+            case transformed
+            when Array
+              consumed = transformed.length
+              result.concat(transformed.compact)
+              i += consumed
+            when nil
+              i += 1
+            else
+              result << transformed
+              i += 1
+            end
+          end
+
+          result
+        end
+
+        # Dispatch a single body element, handling paragraphs specially
+        def dispatch_element(element, index, elements, context)
+          # Paragraphs need style-based dispatch (heading, list, or plain)
+          return dispatch_paragraph(element, index, elements, context) if paragraph?(element)
+
+          # Tables go through registry directly
+          context.transform(element)
+        end
+
+        def dispatch_paragraph(paragraph, index, elements, context)
+          resolver = context.style_resolver
+
+          # Check for section break in paragraph properties
+          if section_break?(paragraph)
+            # Section break without heading → thematic break
+            return section_break_element(paragraph, context)
+          end
+
+          # Heading
+          return @heading_rule.apply(paragraph, context) if resolver.heading?(paragraph)
+
+          # List item — group consecutive items with same numId
+          return group_list(elements, index, context) if resolver.list_item?(paragraph)
+
+          # Regular paragraph (via registry)
+          context.transform(paragraph)
+        end
+
+        def section_break?(paragraph)
+          return false unless paragraph.is_a?(Uniword::Wordprocessingml::Paragraph)
+          return false unless paragraph.properties
+
+          sect_pr = paragraph.properties.section_properties
+          return false unless sect_pr
+
+          sect_pr.type ? true : false
+        end
+
+        def section_break_element(paragraph, context)
+          # First, transform the paragraph content if it has text
+          content = paragraph.runs&.map { |r| r.text&.content.to_s }&.join
+          if content && !content.strip.empty?
+            # Has content — transform normally (content comes before the break)
+            context.transform(paragraph)
+          else
+            # Standalone section break → thematic break
+            Coradoc::CoreModel::Block.new(
+              element_type: 'thematic_break'
+            )
+          end
+        end
+
+        # Collect consecutive list items with the same numId into a ListBlock
+        def group_list(elements, start_index, context)
+          first = elements[start_index]
+          num_id = first.properties&.num_id.to_i
+          items = []
+          consumed = 0
+
+          idx = start_index
+          while idx < elements.length
+            para = elements[idx]
+            break unless paragraph?(para)
+            break unless context.style_resolver.list_item?(para)
+            break unless para.properties&.num_id.to_i == num_id
+
+            items << @list_item_rule.apply(para, context)
+            consumed += 1
+            idx += 1
+          end
+
+          list_block = Coradoc::CoreModel::ListBlock.new(
+            marker_type: context.numbering_resolver.marker_type(num_id),
+            items: items
+          )
+
+          # Return array so caller knows how many elements were consumed
+          consumed > 1 ? [list_block] + Array.new(consumed - 1, nil) : list_block
+        end
+
+        def body_ordered_elements(body)
+          order = body.is_a?(Uniword::Wordprocessingml::Body) ? body.element_order : nil
+          return body.elements if order.nil? || order.empty?
+
+          p_idx = tbl_idx = sdt_idx = 0
+          order.filter_map do |entry|
+            name = entry.is_a?(String) ? entry : entry.name
+            case name
+            when 'p'
+              para = body.paragraphs[p_idx]
+              p_idx += 1
+              para
+            when 'tbl'
+              tbl = body.tables[tbl_idx]
+              tbl_idx += 1
+              tbl
+            when 'sdt'
+              sdt = body.structured_document_tags&.[](sdt_idx)
+              sdt_idx += 1
+              sdt
+            end
+          end
+        end
+
+        def paragraph?(element)
+          defined?(Uniword::Wordprocessingml::Paragraph) &&
+            element.is_a?(Uniword::Wordprocessingml::Paragraph)
+        end
+
+        def collect_footnotes(document)
+          footnotes = {}
+
+          doc_footnotes = document.footnotes
+          if doc_footnotes.is_a?(Hash)
+            doc_footnotes.each do |id, fn|
+              paragraphs = fn.is_a?(Uniword::Wordprocessingml::Footnote) ? fn.paragraphs : fn[:content]
+              footnotes[id.to_s] = Array(paragraphs)
+            end
+          end
+
+          if defined?(Uniword::Wordprocessingml::Footnotes) &&
+             document.footnotes.is_a?(Uniword::Wordprocessingml::Footnotes)
+            document.footnotes.footnotes.each do |fn|
+              id = fn.id&.to_s
+              next unless id
+
+              footnotes[id] = Array(fn.paragraphs || [])
+            end
+          end
+
+          footnotes
+        end
+
+        def extract_document_title(document, context)
+          body = document.body
+          return nil unless body
+
+          paragraphs = body.paragraphs || []
+          paragraphs.each do |para|
+            next unless context.style_resolver.heading?(para)
+            next unless context.style_resolver.heading_level(para) == 1
+
+            runs = para.runs || []
+            return runs.map { |r| r.text&.content.to_s }.join
+          end
+
+          nil
+        end
+
+        def build_registry
+          registry = RuleRegistry.new
+
+          # Only register rules that don't need context for dispatch
+          registry.register(Rules::ParagraphRule.new)
+          registry.register(Rules::RunRule.new)
+          registry.register(Rules::TextRule.new)
+          registry.register(Rules::BreakRule.new)
+          registry.register(Rules::HyperlinkRule.new)
+          registry.register(Rules::ImageRule.new)
+          registry.register(Rules::FootnoteRule.new)
+          registry.register(Rules::BookmarkRule.new)
+          registry.register(Rules::TableRule.new)
+          registry.register(Rules::MathRule.new)
+          registry.register(Rules::StructuredDocumentTagRule.new)
+          registry.register(Rules::SimpleFieldRule.new)
+          registry.register(Rules::ProofErrorRule.new)
+
+          registry
+        end
+
+        # Extract semantic text from headers and footers.
+        # Discards purely layout text ("Page X of Y", page numbers, dates).
+        # Preserves meaningful text (title, version, confidentiality notices).
+        def extract_header_footer_metadata(document, core_doc)
+          extract_from_parts(document, :headers, 'header', core_doc)
+          extract_from_parts(document, :footers, 'footer', core_doc)
+        end
+
+        def extract_from_parts(document, method, prefix, core_doc)
+          parts = case method
+                  when :headers then document.headers
+                  when :footers then document.footers
+                  end
+          return unless parts
+
+          Array(parts).each_with_index do |part, idx|
+            text = extract_part_text(part)
+            next if text.nil? || text.strip.empty?
+            next if layout_only_text?(text)
+
+            part_type = if part.is_a?(Uniword::Wordprocessingml::Header) ||
+                           part.is_a?(Uniword::Wordprocessingml::Footer)
+                          part.type
+                        else
+                          idx
+                        end
+            core_doc.set_metadata("docx.#{prefix}.#{part_type}", text.strip)
+          end
+        end
+
+        def extract_part_text(part)
+          paragraphs = part.paragraphs || []
+          return nil unless paragraphs
+
+          paragraphs.map do |para|
+            extract_paragraph_text_content(para)
+          end.compact.join(' ').strip
+        end
+
+        def extract_paragraph_text_content(para)
+          runs = para.runs || []
+          return nil unless runs
+
+          runs.map { |r| r.text&.content.to_s }.join
+        end
+
+        # Check if header/footer text is purely layout content
+        # (page numbers, "Page X of Y", dates, etc.)
+        def layout_only_text?(text)
+          stripped = text.strip
+          return true if stripped.empty?
+
+          # Pure numbers (page numbers)
+          return true if stripped.match?(/\A\d+\z/)
+
+          # Common page number patterns
+          return true if stripped.match?(/\APage\s+\d+(\s+of\s+\d+)?\z/i)
+
+          # Pure date patterns
+          return true if stripped.match?(%r{\A\d{1,2}[/-]\d{1,2}[/-]\d{2,4}\z})
+          return true if stripped.match?(%r{\A\d{4}[/-]\d{1,2}[/-]\d{1,2}\z})
+
+          false
+        end
+      end
+    end
+  end
+end

data/lib/coradoc/docx/transform.rb

@@ -0,0 +1,38 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Docx
+    module Transform
+      autoload :Rule, 'coradoc/docx/transform/rule'
+      autoload :RuleRegistry, 'coradoc/docx/transform/rule_registry'
+      autoload :Context, 'coradoc/docx/transform/context'
+      autoload :ToCoreModel, 'coradoc/docx/transform/to_core_model'
+      autoload :FromCoreModel, 'coradoc/docx/transform/from_core_model'
+      autoload :StyleResolver, 'coradoc/docx/transform/style_resolver'
+      autoload :NumberingResolver, 'coradoc/docx/transform/numbering_resolver'
+      autoload :OrderedContent, 'coradoc/docx/transform/ordered_content'
+
+      # Element transform rules
+      module Rules
+        autoload :TextRule, 'coradoc/docx/transform/rules/text_rule'
+        autoload :BreakRule, 'coradoc/docx/transform/rules/break_rule'
+        autoload :RunRule, 'coradoc/docx/transform/rules/run_rule'
+        autoload :HyperlinkRule, 'coradoc/docx/transform/rules/hyperlink_rule'
+        autoload :ImageRule, 'coradoc/docx/transform/rules/image_rule'
+        autoload :FootnoteRule, 'coradoc/docx/transform/rules/footnote_rule'
+        autoload :HeadingRule, 'coradoc/docx/transform/rules/heading_rule'
+        autoload :ListItemRule, 'coradoc/docx/transform/rules/list_item_rule'
+        autoload :ParagraphRule, 'coradoc/docx/transform/rules/paragraph_rule'
+        autoload :TableRule, 'coradoc/docx/transform/rules/table_rule'
+        autoload :MathRule, 'coradoc/docx/transform/rules/math_rule'
+        autoload :BookmarkRule, 'coradoc/docx/transform/rules/bookmark_rule'
+        autoload :StructuredDocumentTagRule,
+                 'coradoc/docx/transform/rules/structured_document_tag_rule'
+        autoload :SimpleFieldRule,
+                 'coradoc/docx/transform/rules/simple_field_rule'
+        autoload :ProofErrorRule,
+                 'coradoc/docx/transform/rules/proof_error_rule'
+      end
+    end
+  end
+end

data/lib/coradoc/docx/version.rb

@@ -0,0 +1,7 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module Docx
+    VERSION = '0.1.0'
+  end
+end