Skip to content

Installation

Dependencies

Before we begin, there are several dependencies that PyTNL requires.

Required:

  • Python 3.12 or later, including the development headers for building C/C++ Python modules
  • Compiler for the C++17 standard, e.g. GCC or Clang
  • Git
  • An MPI library such as OpenMPI

Optional:

  • CUDA toolkit for building and using CUDA-enabled PyTNL submodules

Installation

You can install all dependencies with one of the following commands, depending on your Linux distro:

Arch Linux

pacman -S base-devel git python openmpi

Ubuntu

apt install build-essential git python3-dev libopenmpi-dev

Additional dependencies will be pulled in automatically either as Python packages (e.g. cmake) or using the FetchContent cmake module.

PyPi

PyTNL can be installed PyPi using any Python package manager, e.g. pip:

pip install pytnl

Warning

PyTNL currently publishes only a source distribution (sdist) so this step involves building the binary modules on your own system. For this to work, several dependencies must be installed.

Git

The latest development version can be installed directly from the git repository instead of the stable release from PyPI.

pip install git+https://gitlab.com/tnl-project/pytnl.git

Alternatively, if you need to make changes to the sources, see Development.

Warning

This step involves building PyTNL from source, you need to have installed all dependencies

Other

There are other ways to install PyTNL in specific environments, including running plain cmake commands or using a different Python build frontend such as build. See the .gitlab-ci.yml file for examples and do not hesitate to get in touch in case of questions!

Development

This section covers the suggested setup for PyTNL developers. First make sure to install all dependencies.

Clone the PyTNL repository and create a Python virtual environment (venv) for the project:

git clone https://gitlab.com/tnl-project/pytnl.git
cd pytnl
python -m venv .venv
source .venv/bin/activate

Next we need to install the build system in this environment:

pip install scikit-build-core
pip install cmake ninja  # only necessary if not present in your system

To facilitate repeatable builds, the following command installs PyTNL without build isolation using the active venv and shared build subdirectory for build artifacts:

pip install --no-build-isolation -ve .[dev]

Run the previous command again after making changes in the code to rebuild the project.

The [docs] extra installs optional packages for serving and building documentation website.

mkdocs serve # Hosts documentation website locally
mkdocs build # Builds the documentation and outputs static website to /public

The [dev] extra also installs packages for testing and linting the code that you can run:

pytest
ruff check
basedpyright
mypy

The [dev-cuda] extra additionally contains dependencies necessary for testing the CUDA support.