Installation

Supported systems and versions

pyndl currently is only tested and mainly used on 64-bit Linux systems. However, it is possible to install it on other operating systems, but be aware that some functionality might not work or will not work as intended. Therefore be extra careful and run the test suite after installing it on a non Linux system.

Note

If you face problems with installing pyndl with pip, it might be helpful to use Minicoda to install the following dependencies:

conda install numpy cython pandas xarray netCDF4 numpydoc pip

The reason behind this is that during the installation process of pyndl Cython extension need to be installed, if no pre-compiled wheel could be found for your operating system and architecture. To compile Cython extensions some further steps need to be done, which is described in the Cython documentation . These steps depend on your operating system. Installing Cython with conda install cython should add all the necessary additional programs and files and no further steps are needed.

Linux

If you want to install pyndl on Linux the easiest way is to install it from pypi with:

pip install --user pyndl

MacOS

If you want to install pyndl on MacOS you can also install it from pypi. However, the installation will not have openmp support. Sometimes an error is shown during the installation, but the installations succeeds nonetheless. Before filing a bug report please check if you can run the examples from the documentation.

Install pyndl with:

pip install --user pyndl

Windows 10

Note

You might need to enable the bash within Windows 10 first to be able to follow the following instructions.

After installing Anaconda or Miniconda, first install the dependencies with the conda command in the bash or the Miniconda terminal:

conda update conda
conda install numpy cython pandas xarray netCDF4 numpydoc pip

After the installation of the dependencies finished successfully you should be able to install pyndl with pip:

pip install --user pyndl

Warning

This procedure is experimental and might not work. As long as we do not actively support Windows 10 be aware that these installation instructions can fail or the installed package does not always works as intended!