Hylo (formerly Val)

Hylo is a programming language that leverages mutable value semantics and generic programming for high-level systems programming.

This repository contains the sources of the reference implementation of Hylo. Please visit our website to get more information about the language itself.

Developement/Use Requirements

This project is written in Swift and distributed in the form of a package, built with Swift Package Manager. You will need Swift 5.9 or higher to build the compiler from sources.

This repository contains submodules, so after cloning, issue git submodule update --init to populate them.

Windows users: this repository contains symbolic links, so you'll need to enable support for them before checking it out.

LLVM

This package requires LLVM 17. Major versions of LLVM are not interchangeable or backward-compatible.

If you are using this package for development we strongly recommend the use of an LLVM with assertions enabled such as these; otherwise it's much too easy to violate LLVM's preconditions without knowing it. This package's devcontainer (in the .devcontainer subdirectory) has an assert-enabled LLVM build preinstalled in /opt/llvm-MinSizeRel. Instructions to install a debug build are in a comment in .devcontainer/Dockerfile.

If you want to build with the Swift Package Manager and you choose to get LLVM some other way, you'll need an installation with an llvm-config executable, which we will use to create a pkg-config file for LLVM.

Building with CMake and Ninja

1. Configure: choose a build-directory and a CMake build type (usually Debug or Release) and then, where <LLVM> is the path to the root directory of your LLVM installation,

   cmake -D CMAKE_BUILD_TYPE=<build-type> \
     -D LLVM_DIR=<LLVM>/lib/cmake/llvm   \
     -G Ninja -S . -B <build-directory>
   

   (on Windows substitute your shell's line continuation character for \ or just remove the line breaks and backslashes).

   If you want to run tests, add -DBUILD_TESTING=1.

   Note: on macOS, if you are not using your Xcode's default toolchain, you may need -D CMAKE_Swift_COMPILER=swiftc to prevent CMake from using Xcode's default swift.

   If this command fails it could be because you have an LLVM without CMake support installed; we suggest you try one of these packages instead.

2. Build:

   cmake --build <build-directory>
   

3. Test (requires -DBUILD_TESTING=1 in step 1):

   ctest --parallel --test-dir <build-directory>
   

Building with CMake and Xcode

You will need CMake 3.3.0-rc1 or newer.

1. Generate Xcode project: choose a build-directory and then, where <LLVM> is the path to the root directory of your LLVM installation,

   cmake -D LLVM_DIR=<LLVM>/lib/cmake/llvm \
     -G Xcode -S . -B <build-directory>
   

   If you want to run tests, add -DBUILD_TESTING=1.

2. Profit: open the .xcodeproj file in the build-directory and use Xcode's UI to build and test.

Building with Swift Package Manager

Windows Users: Swift Package Manager is poorly supported on Windows and we've been seeing a number of tests act flaky on Windows only when run under Swift Package Manager. We strongly recommend using CMake instead.

Everyone: Swift Package Manager is poorly supported in general and we're considering dropping our use of it, so think about using CMake with ninja or with Xcode instead.

First, you need to create a pkgconfig file specific to your installation and make it visible to your build tools. We use a bash script as follows in the top-level directory of this project:

./Tools/make-pkgconfig.sh ./llvm.pc

if you are on Windows, your git installation (which is required for Swift) contains a bash executable so you can do something like:

C:\Program Files\Git\bin\bash ./Tools/make-pkgconfig.sh ./llvm.pc

The command above generates llvm.pc in the current directory and prints its contents to the terminal. You can either add its directory to your PKG_CONFIG_PATH environment variable for use with command-line tools:

export PKG_CONFIG_PATH=$PWD

or you can put it somewhere that pkg_config already searches (needed for use with Xcode):

sudo mkdir -p /usr/local/lib/pkgconfig && sudo mv llvm.pc /usr/local/lib/pkgconfig/

Once llvm.pc is set up, you should be able to build this project using Swift package manager:

swift build -c release

That command will create an executable named hc in .build/release. That's Hylo's compiler!

To test your compiler,

swift test -c release --parallel

Notes to macOS users:

1. Add platforms: [.macOS("xxx")] to Package.swift where xxx is your macOS version to address the warning complaining that an "object file was built for newer macOS version than being linked".
2. You may need to add the path to zstd library in llvm.pc.

Building a Hylo Devcontainer with VSCode

While Hylo supports Linux natively, it also provides a Devcontainer specification to develop for Linux on other platforms through a Docker container. Our Linux CI uses this specification; this makes it possible to run Linux CI locally on other operating systems like macOS. While this specification should work for any IDE that supports devcontainers, keep in mind this team only uses VSCode.

When opening the Hylo project in VSCode for the first time, you should be prompted to install the extension recommendations in .vscode/extensions.json. If you are not prompted, manually install the extensions by searching for the extension identifiers in the Extensions Marketplace.

Then, build the Devcontainer with the VSCode command: > Dev Containers: Rebuild and Reopen in Container.

Finally, open a new integrated terminal in VSCode and confirm that the shell user is vscode. You can run whoami to check this.

That integrated terminal is connected to the Devcontainer, as if by ssh.
You can now follow the instructions for Building and testing [with CMake and Ninja](#building_with_ cmake_and_ninja) or with Swift Package Manager. All prerequisites, including the llvm.pc file in the default PKG_CONFIG_PATH, are preinstalled.

The Hylo repository files are mounted into the container, so any changes made locally (in VSCode or in other editors) will be automatically propagated into the Devcontainer. However, if you need to modify any of the files in the .devcontainer directory, you will need to rebuild the container with > Dev Containers: Rebuild and Reopen in Container.

Submitting issues

For submitting issue reports you need to manually update the Hylo version. Without this step, hc --version returns unknown.

./Tools/set-hc-version.sh

Implementation status

This project is under active development; expect things to break and APIs to change.

The compiler pipeline is organized as below. You can select how deep the compiler should go through the pipeline with the following options:

• --emit raw-ast: Only parse the input files and output an untyped AST as a JSON file.
• --typecheck: Run the type checker on the input.
• --emit raw-ir: Lower the typed AST into Hylo IR and output the result in a file.
• --emit ir: Run mandatory IR passes and output the result in a file.
• --emit llvm: Transpile the program to LLVM and output LLVM IR.
• --emit intel-asm: Output Intel assembly for all user module(s).
• --emit binary (default): Produce an executable.

For example, hc --emit raw-ast -o main.json main.hylo will parse main.hylo, write the untyped AST in main.json, and exit the pipeline.

A more detailed description of the current implementation status is available on our roadmap page.

Contributing

We welcome contributions to Hylo. Please read through CONTRIBUTING.md for details on how to get started.

You can also get in touch with the community by joining our Slack or one of our Zoom Meetings (ID: 633 2146 3694, Passcode: 409180) on Tuesdays and Thursdays, 13:00-13:30 Pacific time or 22:00-22:30 Central European time.

License

Hylo is distributed under the terms of the Apache-2.0 license. See LICENSE for details.