hydra-cpp

A simple header-only library to quickly define configurations and add hydra support to your cpp projects.

Requirements

• Boost.Python
• Boost.Preprocessor
• yaml-cpp

Getting started

Defining the configuration

Defining a configuration only requires data-type, name, default value and a required flag.

#include <configuration.h>

CONFIGURATION_SCHEMA(
    DBConfig,
    ((std::string,  driver,     "",     true ))
    ((std::string,  user,       "",     true ))
    ((std::string,  pass,       "",     true ))
    ((int,          timeout,    0,      false))
);

Parsing the configuration

The functions to encode/decode the configuration to/from a YAML file are automatically generated, so you can directly use yaml-cpp as

#include <iostream>
#include <yaml-cpp/yaml.h>

void app(const DBConfig& config){
    std::cout << "Driver: "   << config.db.driver << std::endl;
    std::cout << "User: "     << config.db.user   << std::endl;
    std::cout << "Password: " << config.db.pass   << std::endl;
}

int main(){
    YAML::Node node = YAML::LoadFile("config/config.yaml");
    DBConfig config = node.as<DBConfig>();
    app(config);
    return 0;
}

Using with Hydra

A simple Python binding can be used to connect the cpp project to Hydra

void run_app_with_cfg(const std::string& config_str){
    YAML::Node node = YAML::Load(config_str);
    DBConfig config = node.as<DBConfig>();
    app(config);
}

BOOST_PYTHON_MODULE(app)
{
    using namespace boost::python;
    def("run_app_with_cfg", run_app_with_cfg);
}

Then create a simple experiment script in Python as

import hydra
from omegaconf import DictConfig, OmegaConf

# Compiled binding
from app import run_app_with_cfg

@hydra.main(version_base=None, config_path="config", config_name="config")
def my_experiment(cfg : DictConfig) -> None:
    config = OmegaConf.to_object(cfg)

    # Convert YAML to string and pass it to the C++ project
    run_app_with_cfg(str(config))

if __name__ == "__main__":
    my_experiment()

Read hydra documentation for more info about its functionalities.

Running the example

This repo contains a simple example to test the library.
Open it with Visual Studio Code Dev Containers extensions to get a ready-to-use development environment.
You can run the example using the Python-wrapper experiment.py to use all the hydra functionalities or run the main executable.

TODOs

• std container support
• Improve configuration printing
• More testing

⚠️ Warning ⚠️

This library has not been tested properly yet!
Use it at your own risk and open an issue if necessary 😆

License

hydra-cpp is released under the MIT license. See LICENSE for additional details about it.