Installation¶
Linux¶
Anaconda¶
Installation is easiest if you already have a working Anaconda installation.
In this case, simply navigate to the root directory of the project where the
setup script setup.py is and execute python setup.py install.
Non-Anaconda¶
If you don’t have Anaconda, there may be some additional complications. This package depends upon scipy, matplotlib, and h5py, which have some external C-library dependencies. If you don’t have those python packages pre-installed, you may need the following libraries in order to build them when invoking the setup script:
- scipy
- LAPACK (
liblapack-dev) - BLAS (
libopenblas-dev) - Fortran compiler (
gfortran)
- LAPACK (
- h5py
- HDF5 shared library with development headers (1.8.4 or newer, packaged as
libhdf5-devor similar).
- HDF5 shared library with development headers (1.8.4 or newer, packaged as
- matplotlib
- freetype (
libfreetype6-dev)
- freetype (
Once you have that installed you should be able to successfully run
python setup.py install, although I can’t guarantee it. The code
successfully builds on Travis-CI, and I only had to manually instruct it to
install LAPACK, BLAS, a Fortran compiler, and HDF5.
Issues¶
Building seems to work fine with Python 2.7 when starting with a base Anaconda
installation. When building with a minimal Python installation (i.e. in an
environment created by conda create --name new_env python=2.7 or conda
create --name new_env python=3, there seem to be problems building matplotlib.
In Python 3 I get TypeError: unorderable types: str() < int() at the line
reading if self.version < other.version:, and in Python 2 I get:
/usr/bin/ld: cannot find -lnpymath