Kaleidoscope Compiler

This is a refactored version of the Kaleidoscope compiler from the LLVM tutorial. The code has been organized into multiple files for better maintainability.

File Structure

Header Files

• Token.h - Token enum definitions for the lexer
• AST.h - Abstract Syntax Tree node class definitions
• Lexer.h - Lexer function declarations and global state
• Parser.h - Parser function declarations

Source Files

• Lexer.cpp - Lexical analysis implementation
• Parser.cpp - Parser implementation with AST node creation
• main.cpp - Main application logic and handler functions

Build Instructions

Requirements:

• CMake 3.26+（已在本项目中使用）
• C++23 compatible compiler
• LLVM（建议使用 MSYS2 clang64 或官方 LLVM 二进制包）
• Ninja（若不在 PATH，需要显式指定 CMAKE_MAKE_PROGRAM）

推荐（Windows + MSYS2 clang64）

前置：已安装并配置 C:\msys64\clang64。

cmake -S . -B out/build/msys64-clang -G Ninja ^
  -DMSYS_CLANG64_ROOT=C:/msys64/clang64 ^
  -DCMAKE_MAKE_PROGRAM="C:/Program Files (x86)/Microsoft Visual Studio/2022/BuildTools/Common7/IDE/CommonExtensions/Microsoft/CMake/Ninja/ninja.exe"
cmake --build out/build/msys64-clang

说明：

• 会自动使用 C:\msys64\clang64\bin\clang++.exe
• LLVM 会从 C:\msys64\clang64\lib\cmake\llvm 读取
• ZLIB 会从 C:\msys64\clang64 解析

备选（Windows + LLVM 官方二进制包）

前置：已安装到 C:\LLVM，并确保 clang++ 与 llvm-rc 可用。

cmake -S . -B out/build/clang-ninja -G Ninja ^
  -DCMAKE_CXX_COMPILER=C:/LLVM/bin/clang++.exe ^
  -DCMAKE_RC_COMPILER=C:/LLVM/bin/llvm-rc.exe
cmake --build out/build/clang-ninja

WSL2 Fedora + Visual Studio 2026

前置：在 WSL2 Fedora 中安装依赖：

sudo dnf install clang llvm-devel libcxx-devel libcxxabi-devel zlib-devel cmake ninja-build gdb rsync zip

方法 A — VS 2026 打开文件夹（推荐）

1. 在 VS 中选择 File → Open → Folder，打开项目根目录
2. VS 自动识别 CMakePresets.json
3. 在 Target System 下拉框选择 WSL: FedoraLinux-43
4. 在 Configure Preset 下拉框选择 Linux Clang Release (libc++) 或 Linux Clang Debug (libc++)
5. 正常构建即可

方法 B — WSL2 命令行

cmake --preset linux-clang-release
cmake --build --preset linux-clang-release

Usage

Run the executable and enter expressions:

ready> 2+3*4
Parsed a top-level expr.
ready> def foo(x y) x+y*2
Parsed a function definition.
ready>

Architecture

The compiler is structured into distinct phases:

1. Lexical Analysis (Lexer.cpp) - Breaks input into tokens
2. Parsing (Parser.cpp) - Builds Abstract Syntax Tree from tokens
3. Main Loop (main.cpp) - Handles different input types (definitions, expressions, etc.)

This modular design makes it easy to extend with additional features like:

• Code generation to LLVM IR
• Optimization passes
• Additional language constructs

Next Steps

This refactored code is ready for adding LLVM IR generation. The clean separation between lexing, parsing, and AST representation makes it straightforward to add a CodeGen phase that walks the AST and emits LLVM IR.