johndoe6345789
83f1533bce
Extracted from modular repo and reorganized: Compiler Implementation: - 21 compiler source files (frontend, semantic, IR, codegen, runtime) - 15 comprehensive test files (lexer, parser, type checker, backend, etc.) - 9 compiler usage example programs Architecture (5 phases): - Frontend: Lexer, parser, AST generation (lexer.mojo, parser.mojo, ast.mojo) - Semantic: Type system, checking, symbol resolution (3 files) - IR: MLIR code generation (mlir_gen.mojo, mojo_dialect.mojo) - Codegen: LLVM backend, optimization passes (llvm_backend.mojo, optimizer.mojo) - Runtime: Memory mgmt, reflection, async support (3 files) File Organization: - mojo/compiler/src/: Compiler implementation (21 files, 952K) - mojo/compiler/tests/: Test suite (15 files) - mojo/compiler/examples/: Usage examples (9 files) - mojo/samples/: Mojo language examples (37 files, moved from examples/) Documentation: - mojo/CLAUDE.md: Project-level guide - mojo/compiler/CLAUDE.md: Detailed architecture documentation - mojo/compiler/README.md: Quick start guide - mojo/samples/README.md: Example programs guide Status: - Compiler architecture complete (Phase 4) - Full test coverage included - Ready for continued development and integration Files tracked: - 45 new compiler files (21 src + 15 tests + 9 examples) - 1 moved existing directory (examples → samples) - 3 documentation files created - 1 root CLAUDE.md updated Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
Calling Mojo functions from Python
Mojo now has the ability to be called from Python, with Mojo functions representing themselves as if they were native Python functions. When a Python script is run that references a Mojo module, that module is transparently compiled to present an interface that can be used from Python. At present, the interface that is presented to Python must be manually defined in Mojo code.
A full description of this functionality can be found within the Mojo manual, including current known limitations.
These examples show how to perform basic to more advanced use of Mojo code from Python in order to progressively replace hotspots in Python code with fast Mojo. This includes using Mojo to drive calculations on MAX-compatible GPUs.
The two examples of Mojo functions being called from Python consist of:
- hello.py: A literal "hello world" example, where a string is passed in from Python to Mojo, concatenated with a string in Mojo code, and the result passed back to Python to be printed.
- mandelbrot.py: A parallel calculation of the number of iterations to escape in the Mandelbrot set, performed on a MAX-compatible GPU in Mojo. The Python code calls into the Mojo GPU calculation and gets the results back as ASCII art in a string to be printed.
You can run these examples via Pixi:
pixi run hello
pixi run mandelbrot
or directly in a Python virtual environment where the max PyPI package has
been installed:
python hello.py
python mandelbrot.py