Installing Numpy on Mac OS X 10.5.x Leopard

Installing Numpy according to the Scipy wiki page can be a real pain. As Python and GCC are already part of Mac OS X, some prerequisites are already fulfilled, but an official install didn’t do the trick in my case.

When running the advised tests:

python
import numpy
numpy.test('1','10')
import scipy
scipy.test('1','10')

I immediately had problems testing numpy. A little search led me to Macinscience. Here you can download the Scipy Superpack installer for Mac OS X Intel. Note that according to Macinscience it is also possible to download gFortran elsewhere (as I also did) and download the Scipy Superpack installer which doesn’t install gFortran. For both packs hold that these basically are executable SH scripts which downloads the appropriate unofficial files. You can run the superpack script WITH gFortran installer (the other one only has a different name which you should call) through the terminal from your download directory by typing:

sh superpack_2009.03.28.sh

The downloading will start and you will be prompted for your password eventually. After finishing the installation, the Scipy Superpack will be installed. Please note that the Scipy Superpack is regularly updated. When writing this post, the 2009.03.28 was the most recent version. One can find the latest version on this page. Having that said, I again ran the tests:

python
import numpy
numpy.test('1','10')
import scipy
scipy.test('1','10')

I got another error with the numpy test. As also suggested in the first comment on the Macinscience post: a library called nose is required. Although the provider replied that it would be in the package in the future, it wasn’t present (yet?) in my case. This nose library can additionally be downloaded by typing:

sudo easy_install nose

After this my tests passed (note that the scipy test can take some time!).

Getagged , , ,

Geef een reactie

Het e-mailadres wordt niet gepubliceerd. Vereiste velden zijn gemarkeerd met *