fix: parser improvements for v0.0.40

lib/t_ruby.rb

@@ -11,14 +11,14 @@
 
 # Basic components
 require_relative "t_ruby/type_alias_registry"
+require_relative "t_ruby/heredoc_detector"
 require_relative "t_ruby/body_parser"
 require_relative "t_ruby/parser"
 require_relative "t_ruby/union_type_parser"
 require_relative "t_ruby/generic_type_parser"
 require_relative "t_ruby/intersection_type_parser"
 require_relative "t_ruby/type_erasure"
 require_relative "t_ruby/error_handler"
-require_relative "t_ruby/rbs_generator"
 require_relative "t_ruby/declaration_generator"
 require_relative "t_ruby/compiler"
 require_relative "t_ruby/lsp_server"

lib/t_ruby/compiler.rb

@@ -7,16 +7,17 @@ module TRuby
   # \p{L} matches any Unicode letter, \p{N} matches any Unicode number
   IDENTIFIER_CHAR = '[\p{L}\p{N}_]'
   METHOD_NAME_PATTERN = "#{IDENTIFIER_CHAR}+[?!]?".freeze
+  # Visibility modifiers for method definitions
+  VISIBILITY_PATTERN = '(?:(?:private|protected|public)\s+)?'
 
   class Compiler
-    attr_reader :declaration_loader, :use_ir, :optimizer
+    attr_reader :declaration_loader, :optimizer
 
-    def initialize(config = nil, use_ir: true, optimize: true)
+    def initialize(config = nil, optimize: true)
       @config = config || Config.new
-      @use_ir = use_ir
       @optimize = optimize
       @declaration_loader = DeclarationLoader.new
-      @optimizer = IR::Optimizer.new if use_ir && optimize
+      @optimizer = IR::Optimizer.new if optimize
       @type_inferrer = ASTTypeInferrer.new if type_check?
       setup_declaration_paths if @config
     end
@@ -42,16 +43,16 @@ def compile(input_path)
       source = File.read(input_path)
 
       # Parse with IR support
-      parser = Parser.new(source, use_combinator: @use_ir)
-      parse_result = parser.parse
+      parser = Parser.new(source)
+      parser.parse
 
       # Run type checking if enabled
-      if type_check? && @use_ir && parser.ir_program
+      if type_check? && parser.ir_program
         check_types(parser.ir_program, input_path)
       end
 
       # Transform source to Ruby code
-      output = @use_ir ? transform_with_ir(source, parser) : transform_legacy(source, parse_result)
+      output = transform_with_ir(source, parser)
 
       # Compute output path (respects preserve_structure setting)
       output_path = compute_output_path(input_path, @config.ruby_dir, ".rb")
@@ -63,11 +64,7 @@ def compile(input_path)
       if @config.compiler["generate_rbs"]
         rbs_path = compute_output_path(input_path, @config.rbs_dir, ".rbs")
         FileUtils.mkdir_p(File.dirname(rbs_path))
-        if @use_ir && parser.ir_program
-          generate_rbs_from_ir_to_path(rbs_path, parser.ir_program)
-        else
-          generate_rbs_file_to_path(rbs_path, parse_result)
-        end
+        generate_rbs_from_ir_to_path(rbs_path, parser.ir_program)
       end
 
       # Generate .d.trb file if enabled in config (legacy support)
@@ -87,25 +84,17 @@ def compile(input_path)
     def compile_string(source, options = {})
       generate_rbs = options.fetch(:rbs, true)
 
-      parser = Parser.new(source, use_combinator: @use_ir)
-      parse_result = parser.parse
+      parser = Parser.new(source)
+      parser.parse
 
       # Transform source to Ruby code
-      ruby_output = @use_ir ? transform_with_ir(source, parser) : transform_legacy(source, parse_result)
+      ruby_output = transform_with_ir(source, parser)
 
       # Generate RBS if requested
       rbs_output = ""
-      if generate_rbs
-        if @use_ir && parser.ir_program
-          generator = IR::RBSGenerator.new
-          rbs_output = generator.generate(parser.ir_program)
-        else
-          generator = RBSGenerator.new
-          rbs_output = generator.generate(
-            parse_result[:functions] || [],
-            parse_result[:type_aliases] || []
-          )
-        end
+      if generate_rbs && parser.ir_program
+        generator = IR::RBSGenerator.new
+        rbs_output = generator.generate(parser.ir_program)
       end
 
       {
@@ -134,7 +123,7 @@ def compile_to_ir(input_path)
       end
 
       source = File.read(input_path)
-      parser = Parser.new(source, use_combinator: true)
+      parser = Parser.new(source)
       parser.parse
       parser.ir_program
     end
@@ -380,10 +369,10 @@ def setup_declaration_paths
       @declaration_loader.add_search_path("./lib/types")
     end
 
-    # Transform using IR system (new approach)
+    # Transform using IR system
     def transform_with_ir(source, parser)
       ir_program = parser.ir_program
-      return transform_legacy(source, parser.parse) unless ir_program
+      return source unless ir_program
 
       # Run optimization passes if enabled
       if @optimize && @optimizer
@@ -396,33 +385,15 @@ def transform_with_ir(source, parser)
       generator.generate_with_source(ir_program, source)
     end
 
-    # Legacy transformation using TypeErasure (backward compatible)
-    def transform_legacy(source, parse_result)
-      if parse_result[:type] == :success
-        eraser = TypeErasure.new(source)
-        eraser.erase
-      else
-        source
-      end
-    end
-
     # Generate RBS from IR to a specific path
     def generate_rbs_from_ir_to_path(rbs_path, ir_program)
+      return unless ir_program
+
       generator = IR::RBSGenerator.new
       rbs_content = generator.generate(ir_program)
       File.write(rbs_path, rbs_content) unless rbs_content.strip.empty?
     end
 
-    # Legacy RBS generation to a specific path
-    def generate_rbs_file_to_path(rbs_path, parse_result)
-      generator = RBSGenerator.new
-      rbs_content = generator.generate(
-        parse_result[:functions] || [],
-        parse_result[:type_aliases] || []
-      )
-      File.write(rbs_path, rbs_content) unless rbs_content.empty?
-    end
-
     def generate_dtrb_file(input_path, out_dir)
       dtrb_path = compute_output_path(input_path, out_dir, DeclarationGenerator::DECLARATION_EXTENSION)
       FileUtils.mkdir_p(File.dirname(dtrb_path))
@@ -511,7 +482,8 @@ def erase_parameter_types(source)
       result = source.dup
 
       # Match function definitions and remove type annotations from parameters
-      result.gsub!(/^(\s*def\s+#{TRuby::METHOD_NAME_PATTERN}\s*\()([^)]+)(\)\s*)(?::\s*[^\n]+)?(\s*$)/) do |_match|
+      # Also supports visibility modifiers: private def, protected def, public def
+      result.gsub!(/^(\s*#{TRuby::VISIBILITY_PATTERN}def\s+#{TRuby::METHOD_NAME_PATTERN}\s*\()([^)]+)(\)\s*)(?::\s*[^\n]+)?(\s*$)/) do |_match|
         indent = ::Regexp.last_match(1)
         params = ::Regexp.last_match(2)
         close_paren = ::Regexp.last_match(3)
@@ -558,10 +530,13 @@ def remove_param_types(params_str)
       params.join(", ")
     end
 
-    # Clean a single parameter (remove type annotation)
+    # Clean a single parameter (remove type annotation, preserve default value)
     def clean_param(param)
-      # Match: name: Type or name (supports Unicode identifiers)
-      if (match = param.match(/^(#{TRuby::IDENTIFIER_CHAR}+)\s*:/))
+      # Match: name: Type = value (with default value)
+      if (match = param.match(/^(#{TRuby::IDENTIFIER_CHAR}+)\s*:\s*.+?\s*(=\s*.+)$/))
+        "#{match[1]} #{match[2]}"
+      # Match: name: Type (without default value)
+      elsif (match = param.match(/^(#{TRuby::IDENTIFIER_CHAR}+)\s*:/))
         match[1]
       else
         param
@@ -580,11 +555,4 @@ def erase_return_types(source)
       result
     end
   end
-
-  # Legacy Compiler for backward compatibility (no IR)
-  class LegacyCompiler < Compiler
-    def initialize(config)
-      super(config, use_ir: false, optimize: false)
-    end
-  end
 end

lib/t_ruby/heredoc_detector.rb

@@ -0,0 +1,74 @@
+# frozen_string_literal: true
+
+module TRuby
+  # Detects regions that should be skipped during parsing:
+  # - Heredoc content
+  # - Block comments (=begin/=end)
+  class HeredocDetector
+    # Heredoc start patterns:
+    # <<IDENTIFIER, <<-IDENTIFIER, <<~IDENTIFIER
+    # <<'IDENTIFIER', <<"IDENTIFIER"
+    HEREDOC_START_PATTERN = /<<([~-])?(['"]?)(\w+)\2/
+
+    # Detect all skippable ranges in lines (heredocs and block comments)
+    # @param lines [Array<String>] source lines
+    # @return [Array<Range>] content ranges to skip (0-indexed)
+    def self.detect(lines)
+      ranges = []
+      i = 0
+
+      while i < lines.length
+        line = lines[i]
+
+        # Check for =begin block comment
+        if line.strip == "=begin"
+          start_line = i
+          i += 1
+
+          # Find =end
+          while i < lines.length
+            break if lines[i].strip == "=end"
+
+            i += 1
+          end
+
+          # Range covers from =begin to =end (inclusive)
+          ranges << (start_line..i) if i < lines.length
+        # Check for heredoc
+        elsif (match = line.match(HEREDOC_START_PATTERN))
+          delimiter = match[3]
+          squiggly = match[1] == "~"
+          start_line = i
+          i += 1
+
+          # Find closing delimiter
+          while i < lines.length
+            # For squiggly heredoc or dash heredoc, delimiter can be indented
+            # For regular heredoc, delimiter must be at line start
+            if squiggly || match[1] == "-"
+              break if lines[i].strip == delimiter
+            elsif lines[i].chomp == delimiter
+              break
+            end
+            i += 1
+          end
+
+          # Range covers content lines (after start, up to and including end delimiter)
+          ranges << ((start_line + 1)..i) if i < lines.length
+        end
+
+        i += 1
+      end
+
+      ranges
+    end
+
+    # Check if a line index is inside any skippable region
+    # @param line_index [Integer] line index to check
+    # @param heredoc_ranges [Array<Range>] ranges from detect()
+    # @return [Boolean]
+    def self.inside_heredoc?(line_index, heredoc_ranges)
+      heredoc_ranges.any? { |range| range.include?(line_index) }
+    end
+  end
+end

lib/t_ruby/ir.rb

@@ -778,7 +778,8 @@ def build_method(info)
           name: info[:name],
           params: params,
           return_type: info[:return_type] ? parse_type(info[:return_type]) : nil,
-          body: body
+          body: body,
+          visibility: info[:visibility] || :public
         )
       end
 
@@ -1051,7 +1052,8 @@ def visit_method_def(node)
         end
 
         return_type ||= "untyped"
-        emit("def #{node.name}: (#{params}) -> #{return_type}")
+        visibility_prefix = format_visibility(node.visibility)
+        emit("#{visibility_prefix}def #{node.name}: (#{params}) -> #{return_type}")
       end
 
       def visit_class_decl(node)
@@ -1092,6 +1094,17 @@ def visit_instance_variable(node)
       def emit(text)
         @output << (("  " * @indent) + text)
       end
+
+      def format_visibility(visibility)
+        # RBS only supports private visibility, not protected
+        # See: https://github.com/ruby/rbs/issues/579
+        case visibility
+        when :private
+          "private "
+        else
+          ""
+        end
+      end
     end
 
     #==========================================================================

lib/t_ruby/lsp_server.rb

@@ -787,7 +787,7 @@ def generate_semantic_tokens(text)
       lines = text.split("\n")
 
       # Parse the document to get IR
-      parser = Parser.new(text, use_combinator: true)
+      parser = Parser.new(text)
       parse_result = parser.parse
       parser.ir_program