Linux Installation

If using your system’s built-in Python SpacePy we recommend installing dependencies via the package manager. Our recommended (but not required) standalone Python distribution is Anaconda running 64-bit Python 3. Anaconda includes much of the scientific Python stack.

SpacePy ships 64-bit binaries for most modern Linux distributions. They are built with the manylinux2014_x86_64 Docker images and should work on any x86_64 Linux using glibc2.17 or later, e.g. Ubuntu 16.04 and CentOS 7.

Building from source on Linux requires both a C and a Fortran compiler; a recent GCC is recommended (the C compiler is likely included with your distribution). On Debian and Ubuntu:

sudo apt-get install gfortran

Once this is set up, pip install spacepy should Just Work. If you’re installing as a single user (not in a virtual environment) then add the --user flag.

You will also need the NASA CDF library to use pycdf.

If you prefer to install the dependencies some way other than pip, see Dependencies via conda and Dependencies via system packages.

Dependencies via conda

Installation via pip will automatically install most Python dependencies (but not the NASA CDF library). They can also be installed from conda:

conda install numpy python-dateutil scipy matplotlib h5py

Dependencies via system packages

SpacePy usually works with the system Python on Linux. To install dependencies via the package manager on Debian or Ubuntu:

sudo apt-get install python3-dev python3-h5py python3-matplotlib python3-numpy python3-dateutil python3-scipy

For other distributions, check SpacePy Dependencies and install by hand or via your package manager.

To get the dependencies for building documentation:

sudo apt-get install python3-sphinx python3-numpydoc

CDF

The NASA CDF library is included if installing from a binary wheel. If you build SpacePy from source or wish to use a different version of the CDF library, you will need to install it.

It is recommended to install the ncurses library; on Ubuntu and Debian:

sudo apt-get install ncurses-dev

Download the latest CDF library. Choose the file ending in -dist-all.tar.gz from the linux directory. Untar and cd into the resulting directory. Then build:

make OS=linux ENV=gnu CURSES=yes FORTRAN=no UCOPTIONS=-O2 SHARED=yes all

Use CURSES=no if the curses library is not installed. (The distribution-specific directions above will install curses.)

Install:

sudo make install

This will install the library into the default location /usr/local/cdf, where SpacePy can find it. If you choose to install elsewhere, see the CDF documentation, particularly the notes on the CDF_BASE and CDF_LIB environment variables. SpacePy uses these variables to find the library.

Raspberry Pi

SpacePy works on Raspberry Pi, using Raspberry Pi OS in 32-bit or 64-bit flavors. A few tips:

  • No binary wheels are provided, so you will need to build from source.

  • It is highly recommended to install all dependencies (numpy, etc.) via the system package manager apt-get rather than pip, as prebuilt wheels are not generally available and compiling dependencies on the Pi can take a very long time:

    sudo apt-get install gfortran python3-numpy python3-dateutil python3-scipy python3-h5py python3-matplotlib

  • Similarly, use the --no-build-isolation flag to use the system numpy.