This is an experimental project of a C compiler written in Rust. More specifically, the implementation tries to follow the ISO/IEC 9899:2018 standard (i.e., C17) to a certain extent. It's not a production compiler, and probably will never be. The main purpose of this project is to teach myself compiler data structures, language design and optimization techniques.

In order to build this project, you need to install cargo.

Once cargo is installed, make sure that all tests are passing. Doing so guarentees a well working compiler. You can do that by running the following command:

$ cargo test

If all tests pass, then you're good to go. Build the compiler and other tools by running the following command:

$ cargo build --release

Binaries can be found under the target/release/ directory.

Code documentation can be accessed by running the following command:

$ cargo doc --open

There still isn't any other kind of documentation yet, but it's in the radar.

This is a work in progress. There still isn't any way to compile C code.

C17 is a great, challenging language to make a compiler for. It's also true that one can learn a lot by writing a compiler. That being so, C17 seems to be an option that gets the most out of the experience. Just imagine being able to compile whole big projects out there :)

People that went through the journey of writing a C compiler failed almost exclusively. There are just a few examples of successful and compliant C compilers out there. So, that means this project might never have the glory that I wish it'd have. Either way, full compliance is just a hard to achieve dream, and as long as I'm around, I'll be pushing commits to this repository.

This project was formely built with C under the name cci, which I have discontinued in favor of this one. The reason? I was just tired of writing C code, and, in my opinion, Rust is what C reasonably aims to become, but that's a long way. So, why not pick the better tool for the job that's available right now? Also, it's easier to write tests, easier to design APIs, easier to manage dependencies, easier to write documentation. The list goes on.

Instead of using a backend like LLVM or cranelift, I decided I'm going to write a new IR (intermediate representation) and code generation. Given this is not a serious project, but rather a learning experience, I reasoned that building my own IR is more valuable for that end.

This project is licensed under the MIT license. See LICENSE.