Build from source

This Python library uses C++ backend to perform VF3 calculations. Thus, it has to be compiled before it can be used. For most users it will suffice to use precompiled version (see Installation) that is available on PyPi. But, for VF3Py developers, it is desired to modify the C++ part, compile it for Linux and Windows, and push releases.

Build for single user

git clone --recursive https://gitlab.com/knvvv/vf3py.git
cd vf3py
pip install conan
conan profile detect
cd release_assemble
python release_assemble.py

Build release for PyPi

Building VF3Py release includes 3 steps:

  1. Build C++ part for Linux. This will create several so-files in the ./release_assemble/vf3_release directory.

  2. Build C++ part for Windows. This will create several pyd-files, win_dlldeps.json in the ./release_assemble/vf3_release directory and another directory ./release_assemble/vf3_release/win_dlls

  3. Construct overall release as *.tar.gz-package and then upload it.

Linux

First, prepare mamba/conda environments:

mamba create -n py37 python=3.7 -y
mamba activate py37
pip install numpy twine conan
mamba deactivate

mamba create -n py38 python=3.8 -y
mamba activate py38
pip install numpy twine conan
mamba deactivate

mamba create -n py39 python=3.9 -y
mamba activate py39
pip install numpy twine conan
mamba deactivate

mamba create -n py310 python=3.10 -y
mamba activate py310
pip install numpy twine conan
mamba deactivate

mamba create -n py311 python=3.11 -y
mamba activate py311
pip install numpy twine conan
mamba deactivate

mamba create -n py312 python=3.12 -y
mamba activate py312
pip install numpy twine conan
mamba deactivate

Then, C++ part can be built for all these Python versions using the following script:

cd vf3py/release_assemble
python release_assemble.py -full

Note that release_assemble.py creates inner Bash-sessions and calls conda activate ... inside them, so your bash must start up into base conda environment.

Windows

All building is done in MSYS2 environment. After installation, open up MSYS2 MinGW x64 and download prerequisites:

pacman -Syu
pacman -S --needed base-devel mingw-w64-x86_64-toolchain
pacman -S mingw-w64-x86_64-gsl mingw-w64-x86_64-boost mingw-w64-x86_64-cmake mingw-w64-x86_64-pybind11 git

For MSYS2 to have access to Anaconda installed in your Windows, add the following line to C:\msys64\home\*myusername*\.bash_profile file:

eval "$('/c/tools/Anaconda3/Scripts/conda.exe' 'shell.bash' 'hook')"
# Check the path to your conda.exe (same for mamba)

Then, build VF3Py for a single user (see Build for single user).

Create conda envs for Python >= 3.8:

mamba create -n py38 python=3.8 -y
mamba activate py38
pip install numpy twine conan
mamba deactivate

mamba create -n py39 python=3.9 -y
mamba activate py39
pip install numpy twine conan
mamba deactivate

mamba create -n py310 python=3.10 -y
mamba activate py310
pip install numpy twine conan
mamba deactivate

mamba create -n py311 python=3.11 -y
mamba activate py311
pip install numpy twine conan
mamba deactivate

mamba create -n py312 python=3.12 -y
mamba activate py312
pip install numpy twine conan
mamba deactivate

Finally, build C++ part for all these Python versions:

cd vf3py/release_assemble
python release_assemble.py -full -dll-copy
python release_assemble.py -full

Completing and uploading the release

As a result of previous two steps, these files were produced:

  1. ./release_assemble/vf3_release/*.so (in Linux)

  2. ./release_assemble/vf3_release/*.pyd (in Windows)

  3. ./release_assemble/vf3_release/win_dlls/*.dll (in Windows)

  4. ./release_assemble/vf3_release/win_dlldeps.json (in Windows)

Combine them from Linux and Windows machines in a single vf3_release directory.

To prepare and upload the package release, set the new version in the create_pypi_package.py file and do these steps:

python create_pypi_package.py
twine upload dist/*
# For test upload:
# twine upload --repository testpypi dist/*
username: __token__
password: *PyPi - API key*

The latest release of VF3Py can be found here.