From c6e5374d43a731aabe5894e03d7ab7dbae47a6e1 Mon Sep 17 00:00:00 2001
From: ydah <t.yudai92@gmail.com>
Date: Sat, 5 Sep 2026 07:42:11 +0900
Subject: [PATCH] Support structured semantic value tags

---
 lib/lrama/lexer.rb                       |  3 ++-
 lib/lrama/lexer/token/user_code.rb       |  2 +-
 sig/generated/lrama/lexer.rbs            |  2 ++
 spec/lrama/grammar/code_spec.rb          |  5 +++--
 spec/lrama/lexer/token/user_code_spec.rb |  3 +++
 spec/lrama/lexer_spec.rb                 | 19 +++++++++++++++++++
 6 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/lib/lrama/lexer.rb b/lib/lrama/lexer.rb
index 8e0391255..0054f6944 100644
--- a/lib/lrama/lexer.rb
+++ b/lib/lrama/lexer.rb
@@ -63,6 +63,7 @@ class Lexer
       %start
     ).freeze #: Array[String]
     IDENTIFIER_PATTERN = /[a-zA-Z_.][-a-zA-Z0-9_.]*/.freeze #: Regexp
+    TAG_PATTERN = /<[a-zA-Z_][a-zA-Z0-9_]*(?:(?:\.|->)[a-zA-Z_][a-zA-Z0-9_]*)*>/.freeze #: Regexp
 
     # @rbs (GrammarFile grammar_file) -> void
     def initialize(grammar_file)
@@ -126,7 +127,7 @@ def lex_token
         return [@scanner.matched, Lrama::Lexer::Token::Token.new(s_value: @scanner.matched, location: location)]
       when @scanner.scan(/[\?\+\*]/)
         return [@scanner.matched, Lrama::Lexer::Token::Token.new(s_value: @scanner.matched, location: location)]
-      when @scanner.scan(/<\w+>/)
+      when @scanner.scan(TAG_PATTERN)
         return [:TAG, Lrama::Lexer::Token::Tag.new(s_value: @scanner.matched, location: location)]
       when @scanner.scan(/'.'/)
         return [:CHARACTER, Lrama::Lexer::Token::Char.new(s_value: @scanner.matched, location: location)]
diff --git a/lib/lrama/lexer/token/user_code.rb b/lib/lrama/lexer/token/user_code.rb
index 166f04954..e902b5496 100644
--- a/lib/lrama/lexer/token/user_code.rb
+++ b/lib/lrama/lexer/token/user_code.rb
@@ -41,7 +41,7 @@ def scan_reference(scanner)
           if scanner.scan(/
             # $ references
             # It need to wrap an identifier with brackets to use ".-" for identifiers
-            \$(<[a-zA-Z0-9_]+>)?(?:
+            \$(#{Lrama::Lexer::TAG_PATTERN})?(?:
               (\$)                            # $$, $<long>$
             | (\d+)                           # $1, $2, $<long>1
             | ([a-zA-Z_][a-zA-Z0-9_]*)        # $foo, $expr, $<long>program (named reference without brackets)
diff --git a/sig/generated/lrama/lexer.rbs b/sig/generated/lrama/lexer.rbs
index e0e8c97cd..6d46126cb 100644
--- a/sig/generated/lrama/lexer.rbs
+++ b/sig/generated/lrama/lexer.rbs
@@ -24,6 +24,8 @@ module Lrama
 
     IDENTIFIER_PATTERN: Regexp
 
+    TAG_PATTERN: Regexp
+
     # @rbs (GrammarFile grammar_file) -> void
     def initialize: (GrammarFile grammar_file) -> void
 
diff --git a/spec/lrama/grammar/code_spec.rb b/spec/lrama/grammar/code_spec.rb
index 401b00faf..f31121110 100644
--- a/spec/lrama/grammar/code_spec.rb
+++ b/spec/lrama/grammar/code_spec.rb
@@ -165,6 +165,7 @@
             int rule5;
             int rule6;
             int rule7;
+            struct { int integer; } value;
         }
 
         %token <i> keyword_class
@@ -209,7 +210,7 @@
         rule4: expr '+' expr[expr-right] { @1 + @[expr-right]; @0; }
              ;
 
-        rule5: expr '+' expr { $1 + $<integer>3; }
+        rule5: expr '+' expr { $1 + $<value.integer>3; }
              ;
 
         rule6: expr '+' { $<integer>$ = $1; @$ = @1; } expr { $1 + $<integer>4; }
@@ -268,7 +269,7 @@
 
       it "respects explicit tag in a rule" do
         code = grammar.rules.find {|r| r.lhs.id.s_value == "rule5" }
-        expect(code.translated_code(grammar)).to eq(" (yyvsp[-2].expr) + (yyvsp[0].integer); ")
+        expect(code.translated_code(grammar)).to eq(" (yyvsp[-2].expr) + (yyvsp[0].value.integer); ")
       end
 
       context "midrule action exists" do
diff --git a/spec/lrama/lexer/token/user_code_spec.rb b/spec/lrama/lexer/token/user_code_spec.rb
index 5607c6813..6ee14a8d9 100644
--- a/spec/lrama/lexer/token/user_code_spec.rb
+++ b/spec/lrama/lexer/token/user_code_spec.rb
@@ -21,6 +21,9 @@
       references = Lrama::Lexer::Token::UserCode.new(s_value: " $<long>1 ", location: location).references
       expect(references.count).to eq 1
       expect(references[0]).to eq Lrama::Grammar::Reference.new(type: :dollar, number: 1, index: 1, ex_tag: Lrama::Lexer::Token::Tag.new(s_value: "<long>"), first_column: 1, last_column: 9)
+      references = Lrama::Lexer::Token::UserCode.new(s_value: " $<value.integer>1 ", location: location).references
+      expect(references.count).to eq 1
+      expect(references[0]).to eq Lrama::Grammar::Reference.new(type: :dollar, number: 1, index: 1, ex_tag: Lrama::Lexer::Token::Tag.new(s_value: "<value.integer>"), first_column: 1, last_column: 18)
 
       # $foo
       references = Lrama::Lexer::Token::UserCode.new(s_value: " $foo ", location: location).references
diff --git a/spec/lrama/lexer_spec.rb b/spec/lrama/lexer_spec.rb
index 0b37cf9e1..b038b5303 100644
--- a/spec/lrama/lexer_spec.rb
+++ b/spec/lrama/lexer_spec.rb
@@ -392,6 +392,25 @@
         expect(lexer.next_token).to eq([':', token_class::Token.new(s_value: ':')])
       end
     end
+
+    it 'lexes structured semantic value tags' do
+      tags = ["<value>", "<value.integer>", "<value.pointer->integer>"]
+      grammar_file = Lrama::Lexer::GrammarFile.new("tags.y", tags.join(" "))
+      lexer = Lrama::Lexer.new(grammar_file)
+
+      tags.each do |tag|
+        expect(lexer.next_token).to eq([:TAG, token_class::Tag.new(s_value: tag)])
+      end
+    end
+
+    it 'rejects unsupported type and default tags' do
+      ["<int *>", "<std::vector<int>>", "<*>", "<>"].each do |tag|
+        grammar_file = Lrama::Lexer::GrammarFile.new("tags.y", tag)
+        lexer = Lrama::Lexer.new(grammar_file)
+
+        expect { lexer.next_token }.to raise_error(ParseError, /Unexpected token/)
+      end
+    end
   end
 
   context 'unexpected_token.y' do