An minimal NGSolve addon with Python bindings

using the project template

To create your own NGSolve addon project you perfrom the following steps:

1. git-fork this project
2. fill the project with your C++ and Python files
3. adapt CMakeList.txt (addon_name, C++ and Python files)
4. adapt file pyproject.toml, section [project]
5. adapt src/__init__.py file
6. adapt README.md for installation instructions for your addon-project

installing the addon project

Quick install: install the addon package directly with pip from git:

python -m pip install  git+https://github.com/NGSolve/ngsolve-addon-template.git

Alternative needed for self-compiled NGSolve

python -m pip install scikit-build-core pybind11_stubgen toml
python -m pip install --no-build-isolation git+https://github.com/NGSolve/ngsolve-addon-template.git

test it:

python -m ngsolve_addon_template.demos.exploremesh

Step-by-step installation:

simple step-by-step installation using pip:

git clone https://github.com/NGSolve/ngsolve-addon-template.git
cd ngsolve-addon-template
python -m pip install --no-build-isolation .

alternative step-by-step installation using cmake:

git clone https://github.com/NGSolve/ngsolve-addon-template.git
cd ngsolve-addon-template
mkdir build
cd build
cmake ..
make -j4 install

Troubleshooting

Problem

Error in gihub actions when building Linux package:

    ImportError: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.20' not found

Solution

Update pyproject.toml by adding the following line to the [tool.cibuildwheel] section:

[tool.cibuildwheel]
manylinux-x86_64-image = "manylinux_2_28"

Some more NGSolve addons you can find here:

• https://github.com/TUWien-ASC/NGS-myfe (including vs-code instructions)

Optional C++ library

To export a shared C++ library, splitting the addon into a core library and a Python module, you can use the following CMake code. This will install the library, headers, and relocatable CMake files with the Python package. First, create a library with the C++ sources and headers.

add_ngsolve_addon_library(myaddon_core
  PACKAGE myaddon
  EXPORT_NAME core
  SOURCES src/core.cpp
  PUBLIC_HEADERS src/core.hpp)

Then, create a Python module that links to the core library.

add_ngsolve_addon(_pymyaddon src/python.cpp)
target_link_libraries(_pymyaddon PRIVATE myaddon_core)
ngsolve_addon_set_relative_rpath(_pymyaddon ".")

The ngsolve_addon_set_relative_rpath function ensures that the shared library can be found at runtime and is the relative path to the Python module from your addon package at the installation location.

Consumers use find_package(myaddon CONFIG REQUIRED) and link myaddon::core. A myaddon.config Python module can print the CMake directory, as ngsolve.config does. Normal and post-release CMake package versions come from the Python package metadata automatically.