CppInterOp

CppInterOp exposes API from Clang and LLVM in a backward compatibe way. The API support downstream tools that utilize interactive C++ by using the compiler as a service. That is, embed Clang and LLVM as a libraries in their codebases. The API are designed to be minimalistic and aid non-trivial tasks such as language interoperability on the fly. In such scenarios CppInterOp can be used to provide the necessary introspection information to the other side helping the language cross talk.

Installation | Documentation | CppInterOp API Documentation

CppInterOp Introduction

The CppInterOp library provides a minimalist approach for other languages to bridge C++ entities (variables, classes, etc.). This enables interoperability with C++ code, bringing the speed and efficiency of C++ to simpler, more interactive languages like Python.

Incremental Adoption

CppInterOp can be adopted incrementally. While the rest of the framework is the same, a small part of CppInterOp can be utilized. More components may be adopted over time.

Minimalist by design

While the library includes some tricky code, it is designed to be simple and robust (simple function calls, no inheritance, etc.). The goal is to make it as close to the compiler API as possible, and each routine to do just one thing that it was designed for.

Further Enhancing the Dynamic/Automatic bindings in CPPYY

The main use case for CppInterOp is with the CPPYY service. CPPYY is an automatic run-time bindings generator for Python & C++, and supports a wide range of C++ features (e.g., template instantiation). It operates on demand and generates only what is necessary. It requires a compiler (Cling¹ /Clang-REPL²) that can be available during program runtime.

Once CppInterOp is integrated with LLVM's³ Clang-REPL component (that can then be used as a runtime compiler for CPPYY), it will further enhance CPPYY's performance in the following ways:

Simpler codebase: Removal of string parsing logic will lead to a simpler code base.
LLVM Integration: The CppInterOp interfaces will be a part of the LLVM toolchain (as part of Clang-REPL).
Better C++ Support: C++ features such as Partial Template Specialization will be available through CppInterOp.
Fewer Lines of Code: A lot of dependencies and workarounds will be removed, reducing the lines of code required to execute CPPYY.
Well tested interoperability Layer: The CppInterOp interfaces have full unit test coverage.

'Roots' in High Energy Physics research

Besides being developed as a general-purpose library, one of the long-term goals of CppInterOp is to stay backward compatible and be adopted in the High Energy Physics (HEP) field, as it will become an essential part of the Root framework. Over time, parts of the Root framework can be swapped by this API, adding speed and resilience with it.

Build Instructions for Unix Systems

Build instructions for CppInterOp and its dependencies are as follows. CppInterOP can be built with either Cling and Clang-REPL, so instructions will differ slightly depending on which option you would like to build, but should be clear from the section title which instructions to follow.

Clone CppInterOp and cppyy-backend

First clone the CppInterOp repository, as this contains patches that need to be applied to the subsequently cloned llvm-project repo (if building CppInterOP with Clang-REPL)

git clone --depth=1 https://github.com/compiler-research/CppInterOp.git

and clone cppyy-backend repository where we will be installing the CppInterOp library

git clone --depth=1 https://github.com/compiler-research/cppyy-backend.git

Setup Clang-REPL

Clone the 17.x release of the LLVM project repository.

git clone --depth=1 --branch release/17.x https://github.com/llvm/llvm-project.git
cd llvm-project

Get the following patches required for development work.

compgen -G "../CppInterOp/patches/llvm/clang17-*.patch" > /dev/null && find ../CppInterOp/patches/llvm/clang17-*.patch -printf "%f\n" && git apply ../CppInterOp/patches/llvm/clang17-*.patch

Build Clang-REPL

Clang-REPL is an interpreter that CppInterOp works alongside. Build Clang (and Clang-REPL along with it).

mkdir build && cd build
cmake -DLLVM_ENABLE_PROJECTS=clang                  \
                -DLLVM_TARGETS_TO_BUILD="host;NVPTX"          \
                -DCMAKE_BUILD_TYPE=Release                    \
                -DLLVM_ENABLE_ASSERTIONS=ON                   \
                -DLLVM_USE_LINKER=lld                         \
                -DCLANG_ENABLE_STATIC_ANALYZER=OFF            \
                -DCLANG_ENABLE_ARCMT=OFF                      \
                -DCLANG_ENABLE_FORMAT=OFF                     \
                -DCLANG_ENABLE_BOOTSTRAP=OFF                  \
                ../llvm
cmake --build . --target clang clang-repl --parallel $(nproc --all)

Note the 'llvm-project' directory location.

cd ../
export LLVM_DIR=$PWD
cd ../

Build Cling and related dependencies

Besides the Clang-REPL interpreter, CppInterOp also works alongside the Cling interpreter. Cling depends on its own customised version of llvm-project, hosted under the root-project (see the git path below). Use the following build instructions.

git clone --depth=1 https://github.com/root-project/cling.git
git clone --depth=1 -b cling-llvm13 https://github.com/root-project/llvm-project.git
mkdir llvm-project/build
cd llvm-project/build
cmake -DLLVM_ENABLE_PROJECTS=clang                \
    -DLLVM_EXTERNAL_PROJECTS=cling                \
    -DLLVM_EXTERNAL_CLING_SOURCE_DIR=../../cling  \
    -DLLVM_TARGETS_TO_BUILD="host;NVPTX"          \
    -DCMAKE_BUILD_TYPE=Release                    \
    -DLLVM_ENABLE_ASSERTIONS=ON                   \
    -DLLVM_USE_LINKER=lld                         \
    -DCLANG_ENABLE_STATIC_ANALYZER=OFF            \
    -DCLANG_ENABLE_ARCMT=OFF                      \
    -DCLANG_ENABLE_FORMAT=OFF                     \
    -DCLANG_ENABLE_BOOTSTRAP=OFF                  \
    ../llvm
cmake --build . --target clang --parallel $(nproc --all)
cmake --build . --target cling --parallel $(nproc --all)
cmake --build . --target gtest_main --parallel $(nproc --all)

Note the 'llvm-project' directory location.

cd ../
export LLVM_DIR=$PWD
cd ../

Environment variables

Regardless of whether you are building CppInterOP with Cling or Clang-REPL you will need to define the following Envirnoment variables (as they clear for a new session, it is recommended that you also add these to your .bashrc in linux, or your .bash_profile if on MacOS)

export CB_PYTHON_DIR="$PWD/cppyy-backend/python"
export CPPINTEROP_DIR="$CB_PYTHON_DIR/cppyy_backend"
export CPLUS_INCLUDE_PATH="${CPLUS_INCLUDE_PATH}:${LLVM_DIR}/llvm/include:${LLVM_DIR}/clang/include:${LLVM_DIR}/build/include:${LLVM_DIR}/build/tools/clang/include"
export PYTHONPATH=$PYTHONPATH:$CPYCPPYY_DIR:$CB_PYTHON_DIR

If on MacOS you will also need the following envirnoment variable defined

export SDKROOT=`xcrun --show-sdk-path`

Build CppInterOp

Now CppInterOp can be installed

mkdir CppInterOp/build/
cd CppInterOp/build/

Now if you want to build CppInterOp with Clang-REPL then execute the following commands

cmake -DBUILD_SHARED_LIBS=ON -DUSE_CLING=ON -DUSE_REPL=Off -DCling_DIR=$LLVM_DIR/build -DCMAKE_INSTALL_PREFIX=$CPPINTEROP_DIR ..
cmake --build . --target install --parallel $(nproc --all)

If alternatively you would like to install CppInterOp with Cling then execute the following commands

cmake -DBUILD_SHARED_LIBS=ON -DUSE_CLING=ON -DUSE_REPL=Off -DCling_DIR=$LLVM_DIR/build -DCMAKE_INSTALL_PREFIX=$CPPINTEROP_DIR ..
cmake --build . --target install --parallel $(nproc --all)

Testing CppInterOp

To test the built CppInterOp execute the following command in the CppInterOP build folder

cmake --build . --target check-cppinterop --parallel $(nproc --all)

Now go back to the top level directory in which your building CppInterOP

cd ../..

Now you are in a position to install cppyy following the instructions below.

Building and Install cppyy-backend

Clone the repo, build it and copy library files into python/cppyy-backend directory:

git clone --depth=1 https://github.com/compiler-research/cppyy-backend.git
cd cppyy-backend
mkdir -p python/cppyy_backend/lib build 
cd build
cmake -DCppInterOp_DIR=$CPPINTEROP_DIR ..
cmake --build .

If on a linux system now execute the following command

cp libcppyy-backend.so ../python/cppyy_backend/lib/

and if on MacOS execute the following command

cp libcppyy-backend.dylib ../python/cppyy_backend/lib/

Note go back to the top level build directory

cd ../..

Install CPyCppyy

Create virtual environment and activate it:

python3 -m venv .venv
source .venv/bin/activate

git clone --depth=1 https://github.com/compiler-research/CPyCppyy.git
mkdir CPyCppyy/build
cd CPyCppyy/build
cmake ..
cmake --build .

Note down the path to the build directory as CPYCPPYY_DIR:

export CPYCPPYY_DIR=$PWD
cd ../..

Install cppyy

git clone --depth=1 https://github.com/compiler-research/cppyy.git
cd cppyy
python -m pip install --upgrade . --no-deps
cd ..

Run cppyy

Each time you want to run cppyy you need to: Activate the virtual environment

source .venv/bin/activate

Now you can import cppyy in python

python -c "import cppyy"

Run cppyy tests

Follow the steps in Run cppyy. Change to the test directory, make the library files and run pytest:

cd cppyy/test
make all
python -m pip install pytest
python -m pytest -sv

Further Reading: C++ Language Interoperability Layer

Cling is an interpretive Compiler for C++. ↩
Clang-REPL is an interactive C++ interpreter that enables incremental compilation. ↩
LLVM is a Compiler Framework. It is a collection of modular compiler and toolchain technologies. ↩

fsfod/CppInterOp