markdown-merge 7.0.0 → 7.1.3

data/lib/markdown/merge/file_analysis_base.rb

@@ -0,0 +1,766 @@
+# frozen_string_literal: true
+
+require 'digest'
+
+module Markdown
+  module Merge
+    # Base class for file analysis for Markdown files.
+    #
+    # Parses Markdown source code and extracts:
+    # - Top-level block elements (headings, paragraphs, lists, code blocks, etc.)
+    # - Freeze blocks marked with HTML comments
+    # - Structural signatures for matching elements between files
+    #
+    # Subclasses must implement parser-specific methods:
+    # - #parse_document(source) - Parse source and return document node
+    # - #next_sibling(node) - Get next sibling of a node
+    # - #compute_parser_signature(node) - Compute signature for parser-specific nodes
+    # - #node_type_name(type) - Map canonical type names if needed
+    #
+    # Freeze blocks are marked with HTML comments:
+    #   <!-- markdown-merge:freeze -->
+    #   ... content to preserve ...
+    #   <!-- markdown-merge:unfreeze -->
+    #
+    # @example Basic usage (subclass)
+    #   class FileAnalysis < Markdown::Merge::FileAnalysisBase
+    #     def parse_document(source)
+    #       Markly.parse(source, flags: @flags)
+    #     end
+    #
+    #     def next_sibling(node)
+    #       node.next
+    #     end
+    #   end
+    #
+    # @abstract Subclass and implement parser-specific methods
+    class FileAnalysisBase
+      include Ast::Merge::FileAnalyzable
+
+      # Default freeze token for identifying freeze blocks
+      # @return [String]
+      DEFAULT_FREEZE_TOKEN = 'markdown-merge'
+
+      # @return [Object] The root document node
+      attr_reader :document
+
+      # @return [Array] Parse errors if any
+      attr_reader :errors
+
+      # @return [CommentTracker] Comment tracker for this file
+      attr_reader :comment_tracker
+
+      # NOTE: :source is inherited from Ast::Merge::FileAnalyzable
+
+      # Initialize file analysis
+      #
+      # @param source [String] Markdown source code to analyze
+      # @param freeze_token [String] Token for freeze block markers
+      # @param signature_generator [Proc, nil] Custom signature generator
+      def initialize(source, freeze_token: DEFAULT_FREEZE_TOKEN, signature_generator: nil, **parser_options)
+        @source = source
+        # Split by newlines, keeping trailing empty strings (-1)
+        # But remove the final empty string if source ends with newline
+        # (that empty string represents the "line after the last newline" which doesn't exist)
+        @lines = source.split("\n", -1)
+        @lines.pop if @lines.last == '' && source.end_with?("\n")
+        @comment_tracker = CommentTracker.new(@lines)
+
+        @freeze_token = freeze_token
+        @signature_generator = signature_generator
+        @parser_options = parser_options
+        @errors = []
+
+        # Parse the Markdown source - subclasses implement this
+        @document = DebugLogger.time('FileAnalysisBase#parse') do
+          parse_document(source)
+        end
+
+        # Extract and integrate all nodes including freeze blocks
+        @statements = extract_and_integrate_all_nodes
+
+        DebugLogger.debug('FileAnalysisBase initialized', {
+                            signature_generator: signature_generator ? 'custom' : 'default',
+                            document_children: count_children(@document),
+                            statements_count: @statements.size,
+                            freeze_blocks: freeze_blocks.size
+                          })
+      end
+
+      # Parse the source document.
+      #
+      # @abstract Subclasses must implement this method
+      # @param source [String] Markdown source to parse
+      # @return [Object] Root document node
+      def parse_document(source)
+        raise NotImplementedError, "#{self.class} must implement #parse_document"
+      end
+
+      # Get the next sibling of a node.
+      #
+      # Different parsers use different methods (next vs next_sibling).
+      #
+      # @abstract Subclasses must implement this method
+      # @param node [Object] Current node
+      # @return [Object, nil] Next sibling or nil
+      def next_sibling(node)
+        raise NotImplementedError, "#{self.class} must implement #next_sibling"
+      end
+
+      # Check if parse was successful
+      # @return [Boolean]
+      def valid?
+        @errors.empty? && !@document.nil?
+      end
+
+      # Get shared comment capability information for this analysis.
+      #
+      # @return [Object]
+      def comment_capability
+        @comment_capability ||= comment_tracker.augment(owners: []).capability
+      end
+
+      # Describe how Markdown merges currently own and emit comments.
+      #
+      # Standalone HTML comments are source-augmented and emitted through the
+      # shared synthetic comment layer rather than parser-native comment AST.
+      #
+      # @return [Ast::Merge::Comment::SupportStyle]
+      def comment_support_style
+        @comment_support_style ||= shared_comment_support_style(
+          source: :markdown_source,
+          style: :html_comment,
+          read_strategy: :source_augmented_portable_write
+        )
+      end
+
+      # Get all tracked comments converted to shared comment nodes.
+      #
+      # @return [Array]
+      def comment_nodes
+        comment_tracker.comment_nodes
+      end
+
+      # Get a shared comment node at a specific line.
+      #
+      # @param line_num [Integer] 1-based line number
+      # @return [Object, nil]
+      def comment_node_at(line_num)
+        comment_tracker.comment_node_at(line_num)
+      end
+
+      # Get comments in a line range converted to a shared comment region.
+      #
+      # @param range [Range] Range of 1-based line numbers
+      # @param kind [Symbol] Region kind
+      # @param full_line_only [Boolean] Whether to keep only full-line comments
+      # @return [Object]
+      def comment_region_for_range(range, kind:, full_line_only: false)
+        comment_tracker.comment_region_for_range(
+          range,
+          kind: kind,
+          full_line_only: full_line_only
+        )
+      end
+
+      # Build a passive shared comment attachment for an owner.
+      #
+      # @param owner [Object] Structural owner for the attachment
+      # @param options [Hash] Additional metadata / lookup overrides
+      # @return [Object]
+      def comment_attachment_for(owner, **options)
+        augmented_attachment = comment_augmenter(**options).attachment_for(owner)
+
+        shared_comment_attachment_for(
+          owner,
+          tracker_attachment: augmented_attachment || comment_tracker.comment_attachment_for(owner, **options),
+          **options
+        )
+      end
+
+      # @return [Symbol]
+      def comment_attachment_strategy
+        :normalize_tracked_layout_merge
+      end
+
+      def ruleset_logical_owners
+        {
+          link_definition: :preserve_if_referenced
+        }
+      end
+
+      def ruleset_surfaces
+        [
+          { name: :fenced_code_block, selector: :language_tag }
+        ]
+      end
+
+      def ruleset_delegation_policies
+        [
+          { surface_name: :fenced_code_block, strategy: :by_language }
+        ]
+      end
+
+      # Build a passive shared comment augmenter for this analysis.
+      #
+      # @param owners [Array, nil] Owners used for attachment inference
+      # @param options [Hash] Additional augmenter options
+      # @return [Object]
+      def comment_augmenter(owners: nil, **options)
+        comment_tracker.augment(
+          owners: owners || comment_augmenter_default_owners,
+          **options
+        )
+      end
+
+      # Get all statements (block nodes outside freeze blocks + FreezeNode instances)
+      # @return [Array<Object, FreezeNode>]
+      attr_reader :statements
+
+      # Compute default signature for a node
+      # @param node [Object] The parser node or FreezeNode
+      # @return [Array, nil] Signature array
+      def compute_node_signature(node)
+        case node
+        when Ast::Merge::FreezeNodeBase
+          node.signature
+        when LinkDefinitionNode
+          node.signature
+        when GapLineNode
+          node.signature
+        else
+          compute_parser_signature(node)
+        end
+      end
+
+      # Override to detect parser nodes for signature generator fallthrough
+      # @param value [Object] The value to check
+      # @return [Boolean] true if this is a fallthrough node
+      def fallthrough_node?(value)
+        value.is_a?(Ast::Merge::FreezeNodeBase) ||
+          value.is_a?(LinkDefinitionNode) ||
+          value.is_a?(GapLineNode) ||
+          parser_node?(value) ||
+          super
+      end
+
+      # Check if value is a parser-specific node.
+      #
+      # @param value [Object] Value to check
+      # @return [Boolean] true if this is a parser node
+      def parser_node?(value)
+        # Default: check if it responds to :type (common for AST nodes)
+        value.respond_to?(:type)
+      end
+
+      # Compute signature for a parser-specific node.
+      #
+      # @abstract Subclasses should override this method
+      # @param node [Object] The parser node
+      # @return [Array, nil] Signature array
+      def compute_parser_signature(node)
+        type = node.type
+        case type
+        when :heading, :header
+          level = node.header_level
+          # H1 is the document title — treat as a singleton.
+          # A well-formed markdown document has exactly one H1. Matching by text
+          # would cause a generic template title ("AGENTS.md - Development Guide")
+          # and a project-qualified destination title ("AGENTS.md - myGem Development Guide")
+          # to be treated as different nodes, keeping both in the merged output.
+          # Using level-only for H1 makes them the same structural slot so the
+          # preferred version wins cleanly without duplication.
+          return [:heading, 1] if level == 1
+
+          # H2+ match by level and normalized text content
+          [:heading, level, extract_text_content(node)]
+        when :paragraph
+          # Content-based: Match paragraphs by content hash (first 32 chars of digest)
+          text = extract_text_content(node)
+          [:paragraph, Digest::SHA256.hexdigest(text)[0, 32]]
+        when :code_block
+          # Content-based: Match code blocks by fence info and content hash
+          content = safe_string_content(node)
+          fence_info = node.respond_to?(:fence_info) ? node.fence_info : nil
+          [:code_block, fence_info, Digest::SHA256.hexdigest(content)[0, 16]]
+        when :list
+          # Content-fingerprint: Match lists by type and a hash of the first few
+          # items' significant tokens. This lets two lists with similar (but not
+          # identical) content match by signature so item-level inner-merge can run,
+          # rather than the template list being appended as a template-only node.
+          list_type = node.respond_to?(:list_type) ? node.list_type : nil
+          items_text = []
+          child = node.first_child
+          while child
+            items_text << extract_text_content(child).downcase.gsub(/\W+/, ' ').strip
+            child = next_sibling(child)
+            break if items_text.size >= 5
+          end
+          fingerprint = Digest::SHA256.hexdigest(items_text.sort.join('|'))[0, 16]
+          [:list, list_type, fingerprint]
+        when :block_quote, :blockquote
+          # Content-based: Match block quotes by content hash
+          text = extract_text_content(node)
+          [:blockquote, Digest::SHA256.hexdigest(text)[0, 16]]
+        when :thematic_break, :hrule
+          # Structure-based: All thematic breaks are equivalent
+          [:hrule]
+        when :html_block, :html
+          # Content-based: Match HTML blocks by content hash
+          content = safe_string_content(node)
+          [:html, Digest::SHA256.hexdigest(content)[0, 16]]
+        when :table
+          # Content-based: Match tables by structure and header content
+          header_content = extract_table_header_content(node)
+          [:table, count_children(node), Digest::SHA256.hexdigest(header_content)[0, 16]]
+        when :footnote_definition
+          # Name/label-based: Match footnotes by name or label
+          label = node.respond_to?(:name) ? node.name : safe_string_content(node)
+          [:footnote_definition, label]
+        when :custom_block
+          # Content-based: Match custom blocks by content hash
+          text = extract_text_content(node)
+          [:custom_block, Digest::SHA256.hexdigest(text)[0, 16]]
+        else
+          # Unknown type - use type and position
+          pos = node.source_position
+          [:unknown, type, pos&.dig(:start_line)]
+        end
+      end
+
+      # Safely get string content from a node
+      # @param node [Object] The node
+      # @return [String] String content or empty string
+      def safe_string_content(node)
+        node.string_content.to_s
+      rescue TypeError
+        # Some node types don't support string_content
+        extract_text_content(node)
+      end
+
+      # Extract all text content from a node and its children
+      # @param node [Object] The node
+      # @return [String] Concatenated text content
+      def extract_text_content(node)
+        text_parts = []
+        node.walk do |child|
+          if child.type == :text
+            text_parts << child.string_content.to_s
+          elsif child.type == :code
+            text_parts << child.string_content.to_s
+          end
+        end
+        text_parts.join
+      end
+
+      # Get the source text for a range of lines
+      #
+      # Lines are joined with newlines, and each line gets a trailing newline
+      # except for the last line of the file (which may or may not have one in the original).
+      #
+      # @param start_line [Integer] Start line (1-indexed)
+      # @param end_line [Integer] End line (1-indexed)
+      # @return [String] Source text
+      def source_range(start_line, end_line)
+        return '' if start_line < 1 || end_line < start_line
+
+        extracted_lines = @lines[(start_line - 1)..(end_line - 1)]
+        return '' if extracted_lines.empty?
+
+        # Add newlines between and after lines, but not after the last line of the file
+        # unless it originally had one
+        result = extracted_lines.join("\n")
+
+        # Add trailing newline if this isn't the last line of the file
+        # (the last line may or may not have a trailing newline in the original source)
+        if end_line < @lines.length
+          result += "\n"
+        elsif @source&.end_with?("\n")
+          # Last line of file, but original source ends with newline
+          result += "\n"
+        end
+
+        result
+      end
+
+      protected
+
+      # Extract header content from a table node
+      # @param node [Object] The table node
+      # @return [String] Header row content
+      def extract_table_header_content(node)
+        # First row of a table is typically the header
+        first_row = node.first_child
+        return '' unless first_row
+
+        extract_text_content(first_row)
+      end
+
+      # Count children of a node
+      # @param node [Object] The node
+      # @return [Integer] Child count
+      def count_children(node)
+        count = 0
+        child = node.first_child
+        while child
+          count += 1
+          child = next_sibling(child)
+        end
+        count
+      end
+
+      private
+
+      def comment_augmenter_default_owners
+        @comment_augmenter_default_owners ||= @statements.select do |statement|
+          statement.respond_to?(:source_position) && statement.source_position &&
+            (!statement.respond_to?(:merge_type) || statement.merge_type != :gap_line) &&
+            !standalone_comment_statement?(statement)
+        end
+      end
+
+      def standalone_comment_statement?(statement)
+        pos = statement.respond_to?(:source_position) ? statement.source_position : nil
+        return false unless pos
+        return false unless pos[:start_line] && pos[:end_line] && pos[:start_line] == pos[:end_line]
+
+        comment_tracker.comment_node_at(pos[:start_line])
+      end
+
+      # Extract all nodes and integrate freeze blocks
+      # @return [Array<Object>] Integrated list of nodes and freeze blocks
+      def extract_and_integrate_all_nodes
+        freeze_markers = find_freeze_markers
+
+        # Use gap-aware collection to preserve link definitions and gap lines
+        base_nodes = collect_top_level_nodes_with_gaps
+
+        return base_nodes if freeze_markers.empty?
+
+        # Build freeze blocks from markers
+        freeze_blocks = build_freeze_blocks(freeze_markers)
+        return base_nodes if freeze_blocks.empty?
+
+        # Integrate nodes with freeze blocks
+        integrate_nodes_with_freeze_blocks(freeze_blocks, base_nodes)
+      end
+
+      # Collect top-level nodes from document
+      # @return [Array<Object>]
+      def collect_top_level_nodes
+        nodes = []
+        child = @document.first_child
+        while child
+          nodes << child
+          child = next_sibling(child)
+        end
+        nodes
+      end
+
+      # Collect top-level nodes with gap line detection.
+      #
+      # Markdown parsers consume certain content (like link reference definitions)
+      # during parsing. This method detects "gap" lines that aren't covered by any
+      # node and creates synthetic nodes for them.
+      #
+      # @return [Array<Object>] Nodes including gap line nodes
+      def collect_top_level_nodes_with_gaps
+        parser_nodes = collect_top_level_nodes
+        return parser_nodes if @lines.empty?
+
+        # Track which lines are covered by parser nodes
+        covered_lines = Set.new
+        parser_nodes.each do |node|
+          pos = node.source_position
+          next unless pos
+
+          start_line = pos[:start_line]
+          end_line = pos[:end_line]
+
+          # Handle Markly's buggy position reporting for :html nodes
+          # where end_line can be less than start_line (e.g., "lines 3-2")
+          if end_line < start_line
+            # Just mark the start_line as covered
+            covered_lines << start_line
+          else
+            (start_line..end_line).each { |l| covered_lines << l }
+          end
+        end
+
+        # Find gap lines (lines not covered by any node)
+        total_lines = @lines.size
+        gap_line_numbers = (1..total_lines).to_a - covered_lines.to_a
+
+        # Create nodes for gap lines
+        gap_nodes = create_gap_nodes(gap_line_numbers)
+
+        # Integrate gap nodes with parser nodes in line order
+        integrate_gap_nodes(parser_nodes, gap_nodes)
+      end
+
+      # Create nodes for gap lines.
+      #
+      # Link reference definitions get LinkDefinitionNode, others get GapLineNode.
+      # Every gap line gets a node so we can reconstruct the document exactly.
+      #
+      # @param line_numbers [Array<Integer>] Gap line numbers (1-based)
+      # @return [Array<Object>] Gap nodes
+      def create_gap_nodes(line_numbers)
+        line_numbers.map do |line_num|
+          content = @lines[line_num - 1] || ''
+
+          # Try to parse as link definition first
+          link_node = LinkDefinitionNode.parse(content, line_number: line_num)
+          link_node || GapLineNode.new(content, line_number: line_num)
+        end
+      end
+
+      # Integrate gap nodes with parser nodes in line order.
+      # Sets preceding_node for gap lines to enable context-aware signatures.
+      #
+      # @param parser_nodes [Array<Object>] Parser-generated nodes
+      # @param gap_nodes [Array<Object>] Gap line nodes
+      # @return [Array<Object>] All nodes in line order
+      def integrate_gap_nodes(parser_nodes, gap_nodes)
+        all_nodes = parser_nodes + gap_nodes
+
+        # Sort by start line
+        sorted_nodes = all_nodes.sort_by do |node|
+          pos = node.source_position
+          pos ? pos[:start_line] : 0
+        end
+
+        # Set preceding_node for gap lines based on their position in the sorted list
+        # This allows gap lines to have context-aware signatures
+        sorted_nodes.each_with_index do |node, idx|
+          next unless node.is_a?(GapLineNode) && idx > 0
+
+          # Find the previous non-gap-line node (structural node)
+          preceding = sorted_nodes[0...idx].reverse.find { |n| !n.is_a?(GapLineNode) }
+          node.preceding_node = preceding
+          next unless preceding
+
+          node.preceding_signature = begin
+            compute_node_signature(preceding)
+          rescue StandardError
+            nil
+          end
+        end
+
+        sorted_nodes
+      end
+
+      # Find freeze markers from parsed HTML nodes.
+      #
+      # Freeze markers are HTML comments that Markly parses as :html nodes.
+      # By analyzing the parsed nodes (not raw source lines), we automatically ignore
+      # freeze markers that appear inside code blocks (they're part of the code block's
+      # string content, not separate nodes).
+      #
+      # Note: We only support freeze markers as standalone HTML comment nodes.
+      # Freeze markers embedded inside other HTML tags (e.g., `<div><!-- freeze -->text</div>`)
+      # are not detected because they're part of a larger HTML node's content.
+      #
+      # @return [Array<Hash>] Marker information
+      def find_freeze_markers
+        markers = []
+        pattern = Ast::Merge::FreezeNodeBase.pattern_for(:html_comment, @freeze_token)
+
+        return markers unless @document
+
+        # Walk through top-level nodes looking for HTML nodes with freeze markers
+        child = @document.first_child
+        while child
+          node_type = child.type
+
+          # Check HTML nodes for freeze markers
+          # Handle both raw Markly (:html) and TreeHaver normalized ("html_block", :html_block) types
+          if [:html, :html_block, 'html_block', 'html'].include?(node_type)
+            # Try multiple content extraction methods:
+            # 1. string_content (raw Markly/Commonmarker)
+            # 2. to_commonmark on wrapper
+            # 3. inner_node.to_commonmark (TreeHaver Commonmarker wrapper)
+            content = nil
+
+            if child.respond_to?(:string_content)
+              begin
+                content = child.string_content.to_s
+              rescue TypeError
+                # Some nodes don't have string_content
+                content = nil
+              end
+            end
+
+            content = child.to_commonmark.to_s if (content.nil? || content.empty?) && child.respond_to?(:to_commonmark)
+
+            # TreeHaver Commonmarker wrapper stores content in inner_node
+            if (content.nil? || content.empty?) && child.respond_to?(:inner_node)
+              inner = child.inner_node
+              content = inner.to_commonmark.to_s if inner.respond_to?(:to_commonmark)
+            end
+
+            content ||= ''
+            match = content.match(pattern)
+
+            if match
+              pos = child.source_position
+              marker_type = match[1] # "freeze" or "unfreeze"
+              reason = match[2]      # optional reason
+
+              markers << {
+                line: pos ? pos[:start_line] : 0,
+                type: marker_type.to_sym,
+                text: content.strip,
+                reason: reason,
+                node: child # Keep reference to the actual node
+              }
+            end
+          end
+
+          child = next_sibling(child)
+        end
+
+        DebugLogger.debug('Found freeze markers', { count: markers.size })
+        markers
+      end
+
+      # Build freeze blocks from markers
+      # @param markers [Array<Hash>] Marker information
+      # @return [Array<FreezeNode>] Freeze blocks
+      def build_freeze_blocks(markers)
+        blocks = []
+        stack = []
+
+        markers.each do |marker|
+          case marker[:type]
+          when :freeze
+            stack.push(marker)
+          when :unfreeze
+            if stack.any?
+              start_marker = stack.pop
+              blocks << create_freeze_block(start_marker, marker)
+            else
+              DebugLogger.debug('Unmatched unfreeze marker', { line: marker[:line] })
+            end
+          end
+        end
+
+        # Warn about unclosed freeze blocks
+        stack.each do |unclosed|
+          DebugLogger.debug('Unclosed freeze marker', { line: unclosed[:line] })
+        end
+
+        blocks.sort_by(&:start_line)
+      end
+
+      # Create a freeze block from start and end markers.
+      #
+      # Subclasses may override to provide parser-specific FreezeNode subclass.
+      #
+      # @param start_marker [Hash] Start marker info
+      # @param end_marker [Hash] End marker info
+      # @return [FreezeNode]
+      def create_freeze_block(start_marker, end_marker)
+        start_line = start_marker[:line]
+        end_line = end_marker[:line]
+
+        # Content is between the markers (exclusive)
+        content_start = start_line + 1
+        content_end = end_line - 1
+
+        content = if content_start <= content_end
+                    source_range(content_start, content_end)
+                  else
+                    ''
+                  end
+
+        # Parse the content to get nodes (for nested analysis)
+        parsed_nodes = parse_freeze_block_content(content)
+
+        freeze_node_class.new(
+          start_line: start_line,
+          end_line: end_line,
+          content: content,
+          start_marker: start_marker[:text],
+          end_marker: end_marker[:text],
+          nodes: parsed_nodes,
+          reason: start_marker[:reason]
+        )
+      end
+
+      # Returns the FreezeNode class to use.
+      #
+      # Subclasses should override this to return their own FreezeNode class.
+      #
+      # @return [Class] FreezeNode class
+      def freeze_node_class
+        Ast::Merge::FreezeNodeBase
+      end
+
+      # Parse content within a freeze block.
+      #
+      # Subclasses should override this to use their parser.
+      #
+      # @param content [String] Content to parse
+      # @return [Array<Object>] Parsed nodes
+      def parse_freeze_block_content(content)
+        return [] if content.empty?
+
+        begin
+          content_doc = parse_document(content)
+          nodes = []
+          child = content_doc.first_child
+          while child
+            nodes << child
+            child = next_sibling(child)
+          end
+          nodes
+        rescue StandardError => e
+          # simplecov:disable defensive - parser rarely fails on valid markdown subset
+          DebugLogger.debug('Failed to parse freeze block content', { error: e.message })
+          []
+          # simplecov:enable
+        end
+      end
+
+      # Integrate nodes with freeze blocks
+      # @param freeze_blocks [Array<FreezeNode>] Freeze blocks
+      # @param base_nodes [Array<Object>, nil] Base nodes (defaults to collect_top_level_nodes_with_gaps)
+      # @return [Array<Object>] Integrated list
+      def integrate_nodes_with_freeze_blocks(freeze_blocks, base_nodes = nil)
+        result = []
+        freeze_index = 0
+        current_freeze = freeze_blocks[freeze_index]
+
+        top_level_nodes = base_nodes || collect_top_level_nodes_with_gaps
+
+        top_level_nodes.each do |node|
+          node_start = node.source_position&.dig(:start_line) || 0
+          node_end = node.source_position&.dig(:end_line) || node_start
+
+          # Add any freeze blocks that come before this node
+          while current_freeze && current_freeze.start_line < node_start
+            result << current_freeze
+            freeze_index += 1
+            current_freeze = freeze_blocks[freeze_index]
+          end
+
+          # Skip nodes that are inside a freeze block
+          inside_freeze = freeze_blocks.any? do |fb|
+            node_start >= fb.start_line && node_end <= fb.end_line
+          end
+
+          result << node unless inside_freeze
+        end
+
+        # Add remaining freeze blocks
+        while freeze_index < freeze_blocks.size
+          result << freeze_blocks[freeze_index]
+          freeze_index += 1
+        end
+
+        result
+      end
+    end
+  end
+end