[0.2.0] - 2026-07-08

New Contributors

• @take-cheeze made its first contribution in https://github.com/samyeyo/clx/pull/7

Changed

• Replaces NaN tagging based Value to shadow types memory with inline string support (55824b4)
• Added full support for 64 bits integers (55824b4)
• New numeric dispatch helpers (add/sub/mul/div with Int64 fast path) (55824b4)
• Updates CacheSlot for table inline caching (55824b4)
• shadow type Values migration (badb4d2)
• improved native arithmetic with int64_t code emission (badb4d2)
• improved CacheSlot implementation (badb4d2)
• faster function arguments implementation (badb4d2)
• C++ compiler is now the one used to build clx (4d87603)
• use std::from_chars() for integer parsing (fb2c741)
• Refactored LTable to use a new hash table implementation with improved memory management (5409c3b)
• Updated memory allocation and deallocation strategies for hash table entries (5409c3b)
• Removed clx_* prefix in C++ API for better clarity and consistency (5409c3b)
• Updated with new clx C++ API (b06f157)
• Updated codegen to use new clx C++ API (ec4df84)
• clx C++ API now support coroutine management functions (5409c3b)

Added

• Enables assembler ARM64 context switching to replace the broken ucontext shim on Darwin ARM64 (55824b4)
• add ARM64 coroutine context switching via clx_coro_switch(), conditional on APPLE && aarch64 (9a37ebd)
• coroutines implementation now in runtime.cpp (e1afaf1)
• added support for macOS ARM64 context switching (e1afaf1)
• add SIMD helpers for cross-platform performance optimization (b72701a)
• Add clx -o shorthand for --output flag (4d87603)
• Add coroutine.close() support and update documentation for coroutine functions (7a882e5)
• Add tests for coroutine.close() functionality (d3d6594)
• Add coroutine close functionality and refactor coroutine management using new clx C++ API (bc07f91)
• Implement read_line and read_all functions for improved file handling in io module (d8f6fc4)
• Add built-in function lookup and refactor intrinsic calls (44fcbb9)
• Introduced lookup_builtin function to map module and function names to C++ function names (44fcbb9)
• Updated IntrinsicCall structure to use C++ function names instead of integer identifiers (44fcbb9)
• Refactored intrinsic function parsing in the Parser to utilize a map for better readability and maintainability (44fcbb9)
• Enhanced the yields_number function to exclude specific intrinsic functions from number checks (44fcbb9)
• nil-comparison disqualification in function parameters (47e3d00)
• pure numeric array detection before disqualification (47e3d00)
• optimization for simple record tables and function parameter record arrays (47e3d00)
• stack-based traversal of function bodies detecting arithmetic on params, with call-chain propagation (47e3d00)
• adds param_names filter so function parameters don't leak into g_native_numbers (47e3d00)
• enhanced string handling functions to optimize for inline strings (4cb6c2e)
• introduced SIMD optimizations for strings (4cb6c2e)
• Add tests for coroutine.close functionality (d3d6594)
• Add executable permission to shell scripts (2c9f468)
• Add .gitattributes to enforce LF line endings (f008cf6)
• Add robots.txt to allow all user agents (7990e69)

Fixed

• Remove CLX_MUSTTAIL macro definition for Clang compatibility (7a60bd2)
• remove CLX_MUSTTAIL macro definitions from code emitter (a2a56e6)
• enhance compiler detection, path searching, and linking for Apple platforms (7e84049)
• add _XOPEN_SOURCE definition for Apple builds (21dbcad)
• suppress deprecated declaration warnings for Apple builds (1f466fd)
• Fix function call in math_tointeger for consistency with naming conventions (e69d38b)
• Fix GitHub link in footer to correct URL (bcdb0a4)
• fix minifier: drop inline css/js minification (9809d19)

Refactored

• Refactor table functions to improve consistency in version handling (94bc75e)
• Refactor string handling functions for improved consistency and performance (57c94dd)
• Refactor pack_require and luastd_package functions for improved readability and consistency (44d2742)
• Refactor time wrapper functions to remove 'clx_' prefix for consistency (7b527cf)
• Refactor function names to use 'lua_' prefix for consistency in base module (cf24988)
• Remove commented line in test_api function for clarity (a93e67b)
• Remove Bitwise Operations section from index.html (old implementation) (0b365b3)
• Remove bitwise operations section from runtime.md (old implementation) (0f813e7)

Documentation

• updated with new clx ValueTypes aliases (26ef75b)
• updated for clx 0.2.0 (396f269 and 9ff2daa)
• fixed the command to run tests (4a20cae)
• Updated clx C++ API documentation and webpage (1d342bb)

Build

• add ARM64 coroutine assembly for macOS and set defines for current used compiler (6cb22d8)
• update script permissions and add platform-specific linker flags (1f66c72)
• Use NMakefile generator on Windows (a2c60e2)
• Now static CRT is used for clx.exe when using MSVC, libs use dynamic CRT (2296a17)

Benchmarks

• Benchmark batch script now uses only luajit for *_luajit.lua files (00503d0)
• Add --fast option in benchmarks/run.bat (00503d0)
• Increase default value of N in benchmarks for better consistency (0507629)
• Updated benchmarks result (5a51606)

Website

• Publish website (51c5433)
• use python minifier instead of html-minifier-terser (ae5a711)
• version inline next to clx logo (5e7afe4)
• Fixed github link (a5f3ab0)