Refactor and Modernize Java CUP with AST Auto-Generation Support

[EmptyDreams]

Thank you sincerely for your years of dedicated work on the Java CUP project.

This pull request represents a comprehensive modernization of Java CUP—refactoring its architecture, replacing outdated patterns with modern Java practices, and adding significant new capabilities, most notably automatic AST generation directly from .cup grammar files.

⚠️ Important: This is a breaking change. It is not backward-compatible, and users who adopt this version will need to update their code (e.g., class names, symbol handling, build scripts, CLI flags). This is an intentional redesign, not a drop-in upgrade.

I fully understand that such a large, compatibility-breaking change may not align with the current maintenance goals or release strategy of the upstream project. If you decide not to merge it, I will maintain this version independently as a separate fork.

Key changes include:

1. Standardized class naming:
   The original sym class has been renamed to GrammarSymConstants; similarly, classes like parser—which were prone to naming conflicts with user code—have been updated to follow Java’s PascalCase convention.

2. Refactored source location handling:

   • The original -locations command-line option (which toggled whether position information was attached to symbols) has been removed;
   • A unified Location interface is now used as the standard abstraction for source positions;
   • A new -location CLI option allows users to specify a custom Location implementation (defaults to ComplexLocation), offering greater flexibility and type safety.

3. Removed outdated or redundant components:

   • The xmlaction module has been entirely removed;
   • All Simple-series symbols and their corresponding SymbolFactory implementations are deleted;
   • The minijava2 test file has been removed—it shares identical grammar with minijava, differing only in that it includes a simulation runtime, which provides limited value in the current testing context.

4. Overhauled symbol system:

   • The parser-internal symbol class is now named GrammarSymbol, while the user-facing class remains Symbol to clarify responsibilities;
   • Symbol is simplified into an immutable class with cleaner logic;
   • Specialized ComplexSymbol variants for all primitive types (e.g., IntComplexSymbol) eliminate boxing/unboxing overhead;
   • The symbol factory has been completely rewritten to align with the new Symbol and Location design.

5. Modernized collections:
   Legacy collection types such as Stack and Hashtable have been replaced with modern JDK equivalents (e.g., ArrayList, HashMap). Additionally, a new IntArrayStack is introduced to support integer stack operations without boxing.

6. Build system migration:

   • Ant has been fully deprecated in favor of Gradle;
   • Source layout updated: src/java → src/main/java, test grammars moved to src/test/resources;
   • Tests are now executed via Gradle instead of outdated shell scripts, with test cases updated accordingly.

7. New AST auto-generation feature (core addition):

   • AST node classes can now be auto-generated directly from .cup grammar files;
   • A new -ast CLI option enables AST generation and controls node class naming format;
   • To improve readability, a new %namer directive is introduced:
     By default, a distinct AST class is generated for each non-empty production (productions with identical RHS are merged), but the class name is auto-generated and not human-friendly. Users can append %namer MyNodeName to a production to explicitly assign a readable class name to its AST node;
   • Extended grammar syntax now supports:
     • symbol? — optional;
     • symbol* — nullable list;
     • symbol+ — non-empty list;
     • symbol[sep1 sep2 …]+ or symbol[sep1 sep2 …]*:
       Represents a list separated by one or more delimiters (e.g., comma, semicolon, etc., space-separated). Trailing delimiters are controlled by ! (forbidden) or ? (optional). For example:
       • symbol[comma !]+ → non-empty list, no trailing comma;
       • symbol[comma ?]* → nullable list, trailing comma optional;
     • ...symbol — inline (flatten) child node into the current AST node;
   • All AST nodes provide a toTreeString(boolean) method for human-readable tree printing, with console color highlighting when true.

8. Miscellaneous improvements:

   • Full code formatting applied;
   • Unused methods and redundant variables removed;
   • Logic optimized for better performance and readability.

9. And more...

Status:

• ✅ minijava and calc test cases have been updated and pass;
• ⏳ The c language test case is not yet migrated (in progress);
• 📝 Documentation (manual.html, changelog) remains unupdated—will be completed promptly if the PR is accepted.

Thank you again for your time and consideration. Regardless of the outcome, I deeply appreciate the foundation Java CUP has provided—and hope this contribution offers a compelling path forward for those seeking a more modern, expressive, and maintainable parser generator.