Improvement/Modern C++ Wrapper

[Muppetsg2]

Improvement/Modern C++ Wrapper

Motivation / Context

I've spent some time improving the wrapper and adding missing features that were already available in the underlying Tree-sitter C API. This PR aims to make the C++ interface much more comprehensive, safer, and modern, while completely overhauling the CMake build system for better usability.

📌 Important Notes

• Versioning: The version number of this library (currently 0.26.8) now strictly tracks the base version of the underlying Tree-sitter C library. This ensures complete clarity for end-users regarding compatibility.
• Automated Testing: Added support for unit tests using the Catch2 framework to verify the wrapper's behavior. This can be toggled via the CPP_TS_BUILD_TESTS CMake option.
• PR Incorporation: This Pull Request includes and consolidates all changes previously proposed in PR Add getFieldNameForNamedChild #6. (credits to @ashermancinelli)

---

Key Changes

1. Architecture & Compatibility (C++ & OS)

• Multi-Standard Support: Introduced macros and logic to detect the C++ standard in use (TS_HAS_CXX17, TS_HAS_CXX20).
• Backward Compatibility: Implemented a custom ts::StringView fallback to replace std::string_view for pre-C++17 builds.
• Modern C++ (C++20): Utilized concepts for type verification in the new visit() function (with an SFINAE / std::enable_if fallback for older standards).
• Cross-Platform Support: Added OS-specific macros and headers to handle file descriptors correctly across POSIX and Windows (fileno/dup vs _fileno/_dup).
• File Extensions: Renamed the main header from .h to .hpp to better reflect C++ code, along with updated include guards (#pragma once and #ifndef CPP_TREE_SITTER_HPP).

2. Tree-sitter API Expansion

• Query Mechanism: Introduced Tree-sitter's powerful query system with the addition of Query and QueryCursor classes, alongside helper structures like QueryCapture, QueryMatch, QueryCursorState, and QueryPredicateStep.
• Code Editing & Ranges: Added structures to represent source code locations and edits: Point, InputEdit, and Range.
• Input Management: Added an Input struct that enables asynchronous or chunked reading of source text along with encoding support.
• Iterators & Navigation: Introduced the LookaheadIterator class for inspecting possible symbols in a given parser state, and expanded navigation methods within the Node class.
• Visitor Pattern: Implemented a high-level visit() function that allows for easy, depth-first traversal (DFS) of the syntax tree (implemented iteratively for performance).
• Metadata & Scoped Enums: Introduced type-safe scoped enums for constants (InputEncoding, SymbolType, LogType, Quantifier, QueryPredicateStepType) and support for LanguageMetadata.

3. Safety & Memory Management

• Exceptions: Transitioned from silent C-style failures to standard C++ exceptions (std::runtime_error, std::length_error, std::invalid_argument, std::logic_error) for issues like query errors, missing languages, or exceeding 4GB buffer limits.
• Smart Pointers: Expanded the use of std::unique_ptr coupled with custom deleters (details::FreeHelper), ensuring safe memory management for things like query cursors and iterators without leaks.
• Validation: Added a validate() method to InputEdit to ensure edit ranges are mathematically sound before applying them.

4. WebAssembly (WASM) Support

• Wasmtime Integration: Added full support for loading WebAssembly-compiled parsers via the Wasmtime library.
• Environment Management: Introduced the WasmStore class to initialize and manage the lifecycle of the WASM environment and loaded grammars.
• Loading Capabilities: Implemented logic to load WASM grammars directly from memory buffers or .wasm files.

5. CMake Build System Overhaul

• MSVC Support: Added proper compiler flags for Microsoft Visual C++ (/W4, /Zc:__cplusplus) and a CPP_TS_MSVC_STATIC_RUNTIME option to toggle between static (/MT) and dynamic (/MD) MSVC runtime linking.
• Modularization: Split the CMake script into manageable modules - grammar fetching logic is now in grammar.cmake, with helpers in utils.cmake and get_cpm.cmake.
• Enhanced CPPTSAddGrammar: Now searches for grammars in local cache paths (CPP_TS_GRAMMAR_PATH) (resolves Pointing out existing parser's installation path #1).
  • Can download pre-compiled .wasm grammars on-the-fly directly from GitHub Releases.
  • Added support for parsers utilizing C++ scanners (scanner.cc, scanner.cpp), not just C.
• Wasmtime Package Management: Added automatic fetching and linking of the compiled wasmtime library for Windows, macOS, and Linux when the CPP_TS_FEATURE_WASM flag is enabled.
• Amalgamated Builds: Introduced the CPP_TS_AMALGAMATED option to compile Tree-sitter from a single unified file (lib.c), significantly speeding up build times.