Installing

Which Python?

Currently, discretize is tested on python 3.10 through 3.12. We recommend that you use the latest version of Python available on Anaconda.

Installing Python

Python is available on all major operating systems, but if you are getting started with python it is best to use a package manager such as Anaconda. You can download the package manager and use it to install the dependencies above.

Note

When using Continuum Anaconda, make sure to run:

conda update conda
conda update anaconda

Dependencies

discretize’s runtime requirements are:

• numpy 1.22.4 (or greater)

• scipy 1.8 (or greater)

Additional functionality is provided when the following optional packages are installed:

• matplotlib

• pyvista

• vtk

• omf

We also recommend installing:

• pymatsolver 0.1.2 (or greater)

Installing discretize

discretize is available on conda-forge and using the conda (or mamba) package manager is our recommended way of installing discretize`:

conda install -c conda-forge discretize

discretize is also available on pypi:

pip install discretize

There are currently pre-built wheels for windows available on pypi, but other operating systems will require a build.

Installing from Source

Attention

Install discretize from the source code only if you need to run the development version. Otherwise it’s usually better to install it from conda-forge or pypi.

As discretize contains several compiled extensions and is not a pure python pacakge, installing discretize from the source code requires a C/C++ compiler capable of using a C++ 17 standard.

discretize uses a pyproject.toml file to define the build and install steps. As such there is no setup.py file to run. You must use pip to install discretize. As long as you have an available compiler you should be able to install discretize from the source as:

pip install .

Note

For Windows users, make sure you are using compilers that are compatible with your python installation. If you have gnu compilers on your system, meson will default to using those, and you might have to force meson to use the msvc compilers by appending:

--config-settings=setup-args="--vsenv"

to the pip installation command.

Editable Installs

If you are an active developer of discretize, and find yourself modifying the code often, you might want to install it from source, in an editable installation. discretize uses meson-python to build the external modules and install the package. As such, there are a few extra steps to take:

1. Make sure you have the runtime dependencies installed in your environment (see Dependencies listed above). However, you must install numpy >=2.0 when building discretize.

2. You must also install packages needed to build discretize into your environment. You can do so with pip:

   pip install meson-python meson ninja cython setuptools_scm
   

   Or with conda:

   conda install -c conda-forge meson-python meson ninja cython setuptools_scm
   

   This will allow you to use the build backend required by discretize.

3. Finally, you should then be able to perform an editable install using the source code:

   pip install --no-build-isolation --editable .
   

This builds and installs the local directory to your active python environment in an “editable” mode; when source code is changed, you will be able to make use of it immediately. It also builds against the packages installed in your environment instead of creating and isolated environment to build a wheel for the package, which is why we needed to install the build requirements into the environment.

Testing your installation

Head over to the Examples and download and run any of the notebooks or python scripts.