Preparing Mac OS X for Scientific Python

DANSE/reflectometry is not supported on Mac OS X. We do have versions running internally, but preparing your own machine to run them will require some work.

The core set of packages for DANSE/reflectometry are the following:

The following packages are useful for data analysis, but not currently required for any DANSE/reflectometry software.

A bundled suite of packages is available from Enthought for a fee.

Individual packages for OS 10.4+ are available from pythonmac.org. They are hard to find because they are not linked from the pythonmac.org page, and they are no longer actively updated. These are the versions we are running locally.

Python 2.5

OS 10.5 already comes with python 2.5. You can download the latest official version from python.org.

gcc

The OS X install disks provide a development toolkit including the gcc compiler. This must be installed before attempting to build packages from source. There is also a 10.3 backward compatibility SDK for building binaries that will run on OS 10.3+.

gfortran

Some packages depend on Fortran to compile. The Fortran compiler we tested is available from CRAN, the R project package site.

wxPython

wxPython is required for GUI applications only. Binaries for OS 10.3+ are available from the wx download site. Scroll down to the pre-built binaries section. Either unicode or ansi versions should work.

Numpy

Numpy depends on LAPACK/BLAS, which is supplied by Mac veclib. Binaries for numpy are available from source forge. Be sure to select the python 2.5 dmg file and follow the instructions therein. Building Numpy from source can be challenging.

Matplotlib

Matplotlib depends on numpy. This package is probably needed. It is used by some server applications to generate plots for the web, by scripts for producing batch output and by GUI applications for showing interactive plots.

Unfortunately, the matplotlib project does not maintain binaries for OS X. Installation from source will require a number of packages not shipped with a standard OS X install. Follow the installation instructions on the matplotlib distribution site. And good luck!

Scipy

SciPy depends on numpy. Binaries for numpy are available from source forge. Be sure to select the python 2.5 dmg file and follow the instructions therein. Building scipy from source can be very challenging.

Easy install

Easy install is a command line utility for downloading and building python packages. Save ez_setup.py to a temporary directory and open with "Python Launcher (2.5)".

The set of packages available from the Python Cheese Shop is extensive. Once easy_install is available, most of these packages can be installed using "easy_install packagename" from the terminal.

IPython

IPython is an intractive python command line interface which allows you to mix plots and data analysis in your session. To install, use "easy_install ipython". Run it using "ipython -pylab". Now you will be able to use matrix operations and plotting commands directly, much like you would in Matlab(R) or IDL(R).