coradoc-adoc 2.0.9 → 2.0.11

data/lib/coradoc/asciidoc/parser/content.rb

@@ -24,7 +24,7 @@ module Coradoc
         # :zero :one :many
         def text_line(many_breaks = false, unguarded: false, verbatim: false)
           tl = if verbatim
-                 text_any.as(:text)
+                 raw_verbatim_line.as(:text)
                elsif unguarded
                  literal_space? >> text_any.as(:text)
                else
@@ -38,6 +38,15 @@ module Coradoc
           end
         end
 
+        # A verbatim source/listing line: capture every character up to the
+        # next newline without applying any inline substitutions. Source and
+        # listing blocks must round-trip their text untouched — otherwise
+        # sequences like Liquid `{{ var }}` collide with AsciiDoc attribute
+        # references and get rewritten.
+        def raw_verbatim_line
+          (line_ending.absent? >> any).repeat(1)
+        end
+
         def asciidoc_char
           line_start? >> match['*_:+=\-']
         end

data/lib/coradoc/asciidoc/parser/frontmatter_parser.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module AsciiDoc
+    module Parser
+      # AsciiDoc frontmatter extractor.
+      #
+      # Delegates to the shared +Coradoc::CoreModel::FrontmatterBlock::TextSplitter+
+      # — the single source of truth for the `---\n...\n---\n` convention
+      # (DRY). AsciiDoc retains a local parser module for discoverability
+      # and as the seam for any format-specific extensions should they arise
+      # (e.g., recognizing AsciiDoc-style front matter variants).
+      module FrontmatterParser
+        class << self
+          def call(text)
+            Coradoc::CoreModel::FrontmatterBlock::TextSplitter.call(text)
+          end
+        end
+
+        Result = Coradoc::CoreModel::FrontmatterBlock::TextSplitter::Result
+      end
+    end
+  end
+end

data/lib/coradoc/asciidoc/parser/paragraph.rb

@@ -40,9 +40,7 @@ module Coradoc
         # rubocop:enable Style/OptionalBooleanParameter, Style/NumericPredicate
 
         def paragraph
-          (element_id.maybe >>
-            block_title.maybe >>
-            (attribute_list >> newline).maybe >>
+          (block_header >>
             ((paragraph_text_line(0).repeat(1, 1) >>
                    (newline.repeat(1).as(:line_break) | eof?)) |
               (paragraph_text_line(false).repeat(1) >>

data/lib/coradoc/asciidoc/parser/rule_dispatcher.rb

@@ -0,0 +1,158 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module AsciiDoc
+    module Parser
+      # Wraps every parser rule for Parslet memoization.
+      #
+      # Parameterized rules (e.g., `block_style(n_deep, delimiter, repeater)`)
+      # cannot be memoized by Parslet directly because their result depends
+      # on the args. This module aliases each rule to a per-args dispatch
+      # rule so Parslet sees a memoizable, parameterless rule for every
+      # (rule_name, args) combination. Parameterless rules get a single
+      # memoized alias.
+      #
+      # Also warns when a parser method is defined in more than one parser
+      # module — duplicate definitions are usually a refactoring mistake
+      # and produce confusing "last one wins" behavior at include time.
+      #
+      # Invoked once at Parser::Base load time.
+      module RuleDispatcher
+        DISPATCH_CONFIG = {
+          add_dispatch: true,
+          with_params: true
+        }.freeze
+
+        class << self
+          # @param parser_class [Class] Parser::Base or a subclass
+          def apply(parser_class)
+            parser_methods = collect_rule_names
+            warn_on_duplicates(parser_methods)
+            wrap_rules(parser_class, parser_methods)
+          end
+
+          # Per-instance dispatch invoked from the parameterized rule wrappers.
+          # @param parser_instance [Parser::Base]
+          # @param alias_name [Symbol] The alias created at wrap time
+          # @param args [Array]
+          # @param kwargs [Hash]
+          def dispatch(parser_instance, alias_name, *args, **kwargs)
+            cache = dispatch_cache(parser_instance)
+            key = dispatch_key(alias_name, args, kwargs)
+            unless cache.key?(key)
+              rule_name = dispatch_rule_name(alias_name, key)
+              unless parser_instance.respond_to?(rule_name)
+                build_dispatch_rule(parser_instance.class, alias_name,
+                                    rule_name, args, kwargs)
+              end
+              cache[key] = rule_name
+            end
+            parser_instance.public_send(cache[key])
+          end
+
+          private
+
+          # Walk every parser module and collect rule-method names.
+          # `instance_methods(false)` returns only methods defined in the
+          # module itself, not inherited ones — without `false` we would
+          # also wrap Object#send, Kernel#__send__, Module#class, etc.
+          # Wrapping `__send__` in particular causes infinite recursion.
+          # @return [Hash{Symbol => Array<Module>}]
+          def collect_rule_names
+            parser_constants = Parser.constants - %i[Base Cache FixFiles RuleDispatcher]
+            parser_constants.each_with_object({}) do |const, acc|
+              Parser.const_get(const).instance_methods(false).each do |name|
+                acc[name] ||= []
+                acc[name] << const
+              end
+            end
+          end
+
+          def warn_on_duplicates(parser_methods)
+            parser_methods.each do |name, sites|
+              next unless sites.size > 1
+
+              modules = sites.map { |c| Parser.const_get(c) }
+              Coradoc::Logger.warn(
+                "Parser method '#{name}' is defined #{sites.size} times in #{modules.join(', ')}"
+              )
+            end
+          end
+
+          def wrap_rules(parser_class, parser_methods)
+            parser_methods.each_key do |rule_name|
+              params = parser_class.instance_method(rule_name).parameters
+              if dispatch? && params.empty?
+                wrap_nondispatch(parser_class, rule_name)
+              elsif dispatch? && with_params?
+                wrap_dispatch(parser_class, rule_name)
+              end
+            end
+          end
+
+          def wrap_nondispatch(parser_class, rule_name)
+            alias_name = :"alias_nondispatch_#{rule_name}"
+            guard_name = :"alias_nondispatch_rule_guard_#{rule_name}"
+            return if parser_class.method_defined?(guard_name)
+
+            parser_class.class_eval do
+              alias_method alias_name, rule_name
+              define_method(guard_name) {}
+              rule(rule_name) do
+                public_send(alias_name)
+              end
+            end
+          end
+
+          def wrap_dispatch(parser_class, rule_name)
+            alias_name = :"alias_dispatch_#{rule_name}"
+            guard_name = :"alias_dispatch_rule_guard_#{rule_name}"
+            return if parser_class.method_defined?(guard_name)
+
+            parser_class.class_eval do
+              alias_method alias_name, rule_name
+              define_method(guard_name) {}
+              define_method(rule_name) do |*args, **kwargs|
+                RuleDispatcher.dispatch(self, alias_name, *args, **kwargs)
+              end
+            end
+          end
+
+          def dispatch_cache(parser_instance)
+            parser_instance.instance_variable_get(:@_rule_dispatch_cache) ||
+              parser_instance.instance_variable_set(:@_rule_dispatch_cache, {})
+          end
+
+          def dispatch?
+            DISPATCH_CONFIG[:add_dispatch]
+          end
+
+          def with_params?
+            DISPATCH_CONFIG[:with_params]
+          end
+
+          def dispatch_key(alias_name, args, kwargs)
+            [alias_name, args, kwargs.to_a.sort].hash.abs
+          end
+
+          def dispatch_rule_name(alias_name, key)
+            :"#{alias_name}_#{key}"
+          end
+
+          # Build a Parslet memoizable rule that closes over the captured
+          # args and forwards to the original aliased rule. Using Parslet's
+          # class-level `rule()` (not define_method on singleton_class) is
+          # essential — Parslet's memoization, `as()`, and tree building
+          # depend on the rule going through the standard Parslet machinery.
+          def build_dispatch_rule(parser_class, original_alias, rule_name, args, kwargs)
+            parser_class.class_eval do
+              rule(rule_name) do
+                public_send(original_alias, *args, **kwargs)
+              end
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/coradoc/asciidoc/parser/section.rb

@@ -27,9 +27,7 @@ module Coradoc
         def section_block(level = 2)
           return nil if level > 8
 
-          (attribute_list >> newline).maybe >>
-            element_id.maybe >>
-            (attribute_list >> newline).maybe >>
+          block_header >>
             section_title(level).as(:title) >>
             contents.as(:contents).maybe
         end

data/lib/coradoc/asciidoc/parser/table.rb

@@ -41,10 +41,7 @@ module Coradoc
         # - "^e|" centered cell with emphasis style
 
         def table
-          element_id.maybe >>
-            (attribute_list >> newline).maybe >>
-            block_title.maybe >>
-            (attribute_list >> newline).maybe >>
+          block_header >>
             table_start.capture(:table_delim) >>
             line_ending >>
             table_rows.as(:rows) >>

data/lib/coradoc/asciidoc/parser/text.rb

@@ -113,9 +113,7 @@ module Coradoc
         # inline_image is defined in inline.rb for inline images (image:)
 
         def block_image
-          (element_id.maybe >>
-            block_title.maybe >>
-            (attribute_list >> newline).maybe >>
+          (block_header >>
             match('^i') >> str('mage::') >>
             file_path.as(:path) >>
             attribute_list(:attribute_list_macro) >>

data/lib/coradoc/asciidoc/parser.rb

@@ -5,6 +5,7 @@ module Coradoc
     module Parser
       autoload :Base, "#{__dir__}/parser/base"
       autoload :Cache, "#{__dir__}/parser/cache"
+      autoload :FrontmatterParser, "#{__dir__}/parser/frontmatter_parser"
     end
   end
 end

data/lib/coradoc/asciidoc/serializer/serializers/base.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-require 'coradoc/asciidoc/serializer/serialization_context'
+require_relative '../serialization_context'
 
 module Coradoc
   module AsciiDoc

data/lib/coradoc/asciidoc/serializer/serializers/document.rb

@@ -21,6 +21,13 @@ module Coradoc
 
           def serialize_to_adoc
             parts = []
+            # Frontmatter prefix, if present. The raw YAML text is
+            # already canonical (single source of truth is Codec), so we
+            # only wrap it with `---` delimiters.
+            if @model.frontmatter && !@model.frontmatter.strip.empty?
+              parts << "---\n#{@model.frontmatter.chomp}\n---\n\n"
+            end
+
             # Only add leading newline if we have sections but no header
             parts << "\n" if @model.sections && !@model.sections.empty? && !@model.header
 

data/lib/coradoc/asciidoc/serializer/serializers/list/definition.rb

@@ -8,7 +8,9 @@ module Coradoc
           class Definition < Base
             def to_adoc(model, _options = {})
               @model = model
-              content = +"\n"
+              _attrs = @model.attrs.to_adoc(show_empty: false).to_s
+              prefix = _attrs.empty? ? '' : "#{_attrs}\n"
+              content = "\n#{prefix}"
               @model.items.each do |item|
                 # Pass delimiter to item serialization
                 serialized = serialize_child_with_options(item, delimiter: @model.delimiter)

data/lib/coradoc/asciidoc/transform/callout_merger.rb

@@ -0,0 +1,94 @@
+# frozen_string_literal: true
+
+module Coradoc
+  module AsciiDoc
+    module Transform
+      # Post-processing pass that merges AsciiDoc callout annotation
+      # paragraphs into the verbatim block they annotate.
+      #
+      # AsciiDoc callouts look like:
+      #
+      #   [source,ruby]
+      #   ----
+      #   get '/hi' do <1>
+      #   ----
+      #   <1> Returns hello world
+      #
+      # The parser emits the source block and the annotation as two
+      # independent children. The CoreModel representation should attach
+      # the annotation to the block as a typed Callout, so downstream
+      # serializers can render them appropriately for each format.
+      #
+      # Single responsibility: take a flat list of transformed CoreModel
+      # children, return a flat list with `<N>` paragraphs adjacent to a
+      # SourceBlock / ListingBlock folded into that block's `callouts`.
+      # Anything else is passed through untouched.
+      class CalloutMerger
+        ANNOTATION_LINE = /<(\d+)>\s*(.*?)\s*\z/
+        ANNOTATION_SPLIT = /(?=<\d+>)/
+
+        class << self
+          def call(children)
+            new.merge(Array(children))
+          end
+        end
+
+        # Walks the input children left-to-right. When a paragraph whose
+        # content is composed entirely of `<N> text` lines follows a
+        # verbatim block (SourceBlock or ListingBlock), each `<N>` line
+        # becomes a Callout attached to that block instead of a separate
+        # paragraph.
+        #
+        # Annotations that do not follow a verbatim block are preserved
+        # verbatim — they may be legitimate prose.
+        def merge(children)
+          children.each.with_object([]) do |child, result|
+            annotations = extract_annotations(child)
+            target = annotations && preceding_verbatim_block(result)
+            if target
+              target.callouts.concat(annotations)
+            else
+              result << child
+            end
+          end
+        end
+
+        private
+
+        # Returns an Array of Callout if the paragraph is entirely
+        # callout annotations, otherwise nil.
+        def extract_annotations(child)
+          return nil unless child.is_a?(Coradoc::CoreModel::ParagraphBlock)
+
+          lines = split_annotation_lines(child.flat_text)
+          return nil if lines.empty?
+
+          callouts = lines.map do |line|
+            match = line.match(ANNOTATION_LINE)
+            match ? Coradoc::CoreModel::Callout.new(index: match[1].to_i, content: match[2]) : nil
+          end
+          return nil if callouts.any?(&:nil?)
+
+          callouts
+        end
+
+        # Splits a paragraph into candidate annotation lines. Returns []
+        # if any non-blank segment fails to look like an annotation.
+        def split_annotation_lines(text)
+          return [] if text.nil? || text.strip.empty?
+
+          chunks = text.strip.split(ANNOTATION_SPLIT)
+          chunks.map(&:strip).reject(&:empty?)
+        end
+
+        def preceding_verbatim_block(result)
+          last = result.last
+          return nil unless last.is_a?(Coradoc::CoreModel::SourceBlock) ||
+                            last.is_a?(Coradoc::CoreModel::ListingBlock)
+
+          last
+        end
+      end
+    end
+  end
+end

data/lib/coradoc/asciidoc/transform/element_transformers/block_transformer.rb

@@ -62,7 +62,16 @@ module Coradoc
               has_nested_blocks = lines.any?(Coradoc::AsciiDoc::Model::Block::Core)
 
               if has_nested_blocks
-                children = lines.map { |line| ToCoreModel.transform(line) }
+                children = lines.filter_map do |line|
+                  result = ToCoreModel.transform(line)
+                  next nil if result.nil?
+                  next result if result.is_a?(Coradoc::CoreModel::Base)
+
+                  text = ToCoreModel.extract_text_content(result)
+                  next nil if text.nil? || text.strip.empty?
+
+                  Coradoc::CoreModel::TextContent.new(text: text)
+                end
                 klass.new(
                   id: block.id,
                   title: ToCoreModel.extract_title_text(block.title),

data/lib/coradoc/asciidoc/transform/element_transformers/document_transformer.rb

@@ -9,14 +9,29 @@ module Coradoc
             def transform_document(doc)
               title_text = ToCoreModel.extract_title_text(doc.header&.title)
               attributes = ToCoreModel.extract_document_attributes(doc)
+              children = ToCoreModel.transform(doc.sections || doc.contents || [])
+              children = CalloutMerger.call(children)
+              children = prepend_frontmatter(children, doc.frontmatter)
+
               Coradoc::CoreModel::DocumentElement.new(
                 id: doc.id,
                 title: title_text,
                 attributes: attributes,
-                children: ToCoreModel.transform(doc.sections || doc.contents || [])
+                children: children
               )
             end
 
+            # If the AsciiDoc document carried raw frontmatter text, parse
+            # it via Codec into a typed FrontmatterBlock and prepend it so
+            # it participates in the standard block pipeline. Codec is the
+            # single source of truth for YAML parsing (DRY/MECE).
+            def prepend_frontmatter(children, frontmatter_text)
+              return children if frontmatter_text.nil? || frontmatter_text.strip.empty?
+
+              block = Coradoc::CoreModel::FrontmatterBlock::Codec.from_yaml(frontmatter_text)
+              [block, *children]
+            end
+
             def transform_section(section, parent_id: nil)
               title_text = ToCoreModel.extract_title_text(section.title)
               section_id = section.id || Coradoc::CoreModel::IdGenerator.generate_from_title(
@@ -24,6 +39,7 @@ module Coradoc
               )
 
               content_children = ToCoreModel.transform(section.contents || [])
+              content_children = CalloutMerger.call(content_children)
               nested_sections = (section.sections || []).map do |child|
                 transform_section(child, parent_id: section_id)
               end

data/lib/coradoc/asciidoc/transform/element_transformers/other_transformer.rb

@@ -25,8 +25,10 @@ module Coradoc
             end
 
             def transform_image(image)
+              src = image.src.to_s
+              src = src[1..] if src.start_with?(':')
               Coradoc::CoreModel::Image.new(
-                src: image.src,
+                src: src,
                 alt: image.title&.to_s,
                 width: image.attributes&.[]('width'),
                 height: image.attributes&.[]('height')

data/lib/coradoc/asciidoc/transform/from_core_model.rb

@@ -1,7 +1,5 @@
 # frozen_string_literal: true
 
-require_relative 'from_core_model_registrations'
-
 module Coradoc
   module AsciiDoc
     module Transform
@@ -9,8 +7,18 @@ module Coradoc
       class FromCoreModel
         include Coradoc::Transform::Base
 
+        @registered = false
+
         class << self
+          def register!
+            return if @registered
+
+            Transform::FromCoreModelRegistrations.register_all!
+            @registered = true
+          end
+
           def transform(model)
+            register!
             return model.map { |item| transform(item) } if model.is_a?(Array)
             return model unless model.is_a?(Coradoc::CoreModel::Base)
 
@@ -32,27 +40,42 @@ module Coradoc
                          Coradoc::AsciiDoc::Model::Header.new(title: '')
                        end
 
+              sections, frontmatter = extract_frontmatter(Array(element.children))
+
               Coradoc::AsciiDoc::Model::Document.new(
                 id: element.id,
                 header: header,
-                sections: transform(element.children)
+                sections: flatten_children(sections),
+                frontmatter: frontmatter
               )
             when CoreModel::SectionElement
               Coradoc::AsciiDoc::Model::Section.new(
                 id: element.id,
                 level: element.level,
                 title: create_title(element.title, element.level),
-                contents: transform(element.children)
+                contents: flatten_children(element.children)
               )
             else
               Coradoc::AsciiDoc::Model::Section.new(
                 id: element.id,
                 title: create_title(element.title, 1),
-                contents: transform(element.children)
+                contents: flatten_children(element.children)
               )
             end
           end
 
+          # Transforms each CoreModel child and flattens one level so a
+          # transform that returns multiple siblings (e.g. a source block
+          # followed by its re-expanded callout paragraphs) stays in
+          # document order.
+          def flatten_children(children)
+            Array(children).flat_map { |child| flatten_one(transform(child)) }
+          end
+
+          def flatten_one(result)
+            result.is_a?(Array) ? result : [result]
+          end
+
           def transform_block(block)
             content = block.renderable_content
 
@@ -71,12 +94,19 @@ module Coradoc
             end
 
             content_text = safe_content_to_string(content)
+            result = build_verbatim_block(semantic, block, content_text)
+            return result unless verbatim_with_callouts?(semantic, block)
 
+            [result, *build_callout_paragraphs(block.callouts)]
+          end
+
+          def build_verbatim_block(semantic, block, content_text)
             case semantic
             when :source_code
               Coradoc::AsciiDoc::Model::Block::SourceCode.new(
                 id: block.id,
                 title: block.title,
+                lang: block.language,
                 lines: content_text.split("\n"),
                 attributes: build_attributes(block)
               )
@@ -151,6 +181,23 @@ module Coradoc
             end
           end
 
+          def verbatim_with_callouts?(semantic, block)
+            return false unless %i[source_code listing].include?(semantic)
+            return false if block.callouts.nil? || block.callouts.empty?
+
+            true
+          end
+
+          # Re-expands typed Callouts back into the AsciiDoc `<N> text`
+          # paragraph form so the round-trip is faithful.
+          def build_callout_paragraphs(callouts)
+            callouts.sort_by { |c| c.index || Float::INFINITY }.map do |callout|
+              Coradoc::AsciiDoc::Model::Paragraph.new(
+                content: create_text_elements("<#{callout.index}> #{callout.content}")
+              )
+            end
+          end
+
           def transform_table(table)
             rows = Array(table.rows).map do |row|
               columns = Array(row.cells).map do |cell|
@@ -335,8 +382,26 @@ module Coradoc
             )
           end
 
+          def transform_comment_line(comment)
+            Coradoc::AsciiDoc::Model::CommentLine.new(
+              text: comment.text.to_s
+            )
+          end
+
           private
 
+          # If the first CoreModel child is a FrontmatterBlock, serialize
+          # it to YAML text via Codec (single source of truth) and pop it
+          # from the children list. Returns [remaining_children,
+          # frontmatter_text].
+          def extract_frontmatter(children)
+            first = children.first
+            return [children, nil] unless first.is_a?(CoreModel::FrontmatterBlock)
+
+            yaml = CoreModel::FrontmatterBlock::Codec.to_yaml(first)
+            [children.drop(1), yaml.nil? || yaml.empty? ? nil : yaml]
+          end
+
           def resolve_semantic_type(block)
             semantic = block.resolve_semantic_type
             return semantic if semantic

data/lib/coradoc/asciidoc/transform/from_core_model_registrations.rb

@@ -115,6 +115,11 @@ module Coradoc
               Coradoc::CoreModel::BibliographyEntry,
               ->(model) { FromCoreModel.transform_bibliography_entry(model) }
             )
+
+            Registry.register(
+              Coradoc::CoreModel::CommentLine,
+              ->(model) { FromCoreModel.transform_comment_line(model) }
+            )
           end
         end
       end
@@ -123,4 +128,3 @@ module Coradoc
 end
 
 # Auto-register when this file is loaded
-Coradoc::AsciiDoc::Transform::FromCoreModelRegistrations.register_all!