pybind11_json is an nlohmann::json to pybind11 bridge, it allows you to automatically convert nlohmann::json to py::object and the other way around. Simply include the header, and the automatic conversion will be enabled.
Tip
Looking for an equivalent with nanobind? Check this out https://github.com/ianhbell/nanobind_json !
#include "pybind11_json/pybind11_json.hpp"
#include "nlohmann/json.hpp"
#include "pybind11/pybind11.h"
namespace py = pybind11;
namespace nl = nlohmann;
using namespace pybind11::literals;
py::dict obj = py::dict("number"_a=1234, "hello"_a="world");
// Automatic py::dict->nl::json conversion
nl::json j = obj;
// Automatic nl::json->py::object conversion
py::object result1 = j;
// Automatic nl::json->py::dict conversion
py::dict result2 = j;
You can easily make bindings for C++ classes/functions that make use of nlohmann::json.
For example, making a binding for the following two functions is automatic, thanks to pybind11_json
void take_json(const nlohmann::json& json) {
std::cout << "This function took an nlohmann::json instance as argument: " << s << std::endl;
}
nlohmann::json return_json() {
nlohmann::json j = {{"value", 1}};
std::cout << "This function returns an nlohmann::json instance: " << j << std::endl;
return j;
}
Bindings:
PYBIND11_MODULE(my_module, m) {
m.doc() = "My awesome module";
m.def("take_json", &take_json, "pass py::object to a C++ function that takes an nlohmann::json");
m.def("return_json", &return_json, "return py::object from a C++ function that returns an nlohmann::json");
}
You can then use your functions Python side:
import my_module
my_module.take_json({"value": 2})
j = my_module.return_json()
print(j)
You can find an example of simple Python bindings using pybind11_json here: https://github.com/martinRenou/xjson
You can install pybind11_json using conda
conda install -c conda-forge pybind11 nlohmann_json pybind11_json
We encourage you to use conda for installing dependencies, but it is not a requirement for pybind11_json to work
conda install cmake nlohmann_json pybind11 -c conda-forge
Then you can install the sources
cmake -D CMAKE_INSTALL_PREFIX=your_conda_path
make install
Download the "pybind11_json.hpp" single file into your project, and install/download pybind11 and nlohmann_json or use as git submodule.
You can compile and run tests locally doing
cmake -D CMAKE_INSTALL_PREFIX=$CONDA_PREFIX -D DOWNLOAD_GTEST=ON ..
make
./test/test_pybind11_json
pybind11_json depends on
pybind11_json |
nlohmann_json |
pybind11 |
|---|---|---|
| master | >=3.2.0,<4.0 | >=2.2.4,<3.0 |
| 0.2.* | >=3.2.0,<4.0 | >=2.2.4,<3.0 |
We use a shared copyright model that enables all contributors to maintain the copyright on their contributions.
This software is licensed under the BSD-3-Clause license. See the LICENSE file for details.