Okay, I am well in the third day of trying to install SciPy on the Mac OS X, and things do not seem they are improving:
- I have started by signing on the Apple Developer Program, downloading and installing Xcode (without which nothing else is possible);
- I’ve downloaded and installed MacPorts (without which nothing interesting is humanely feasible);
- As required, I’ve installed a fresh copy of Python 2.6 from the .dmg, available at python.org;
- I’ve installed gfortran from the binary, available at AT&T Research;
- Using MacPorts, I’ve installed the SuiteSparse;
- I’ve installed numpy from the .dmg available at SourceForge;
- I’ve installed the setuputils for python, downloading the egg and running it as a shell script, as indicated on its website;
- Using setuputils, I’ve installed nose, the testing framework, and run the smoke test for numpy (
import numpy \n numpy.test('1', '10')). It passed perfectly;
- I’ve installed scipy from the .dmg available at SourceForge;
- I’ve run the smoke test for scipy (
import numpy \n scipy.test('1', '10')). It smoked badly at all attempts of code weaving with C(++).
This short version of the story does not include all the comings and goings — installing Python 2.7 and discovering very late in the process its incompatibility with Ccipy for Mac, finding out that the concept of automatic uninstallation simply does not exist in Mac OS X, it has been lots of fun ! It feels just like UNIX, only worse — I have fallen in the rabbit hole and I am waiting for the floor to arrive.
I swear that I am this close to just installing Windows on this damn thing.
EDIT 4/feb: I have just found out this very useful script, Scipy Superpack for Mac, courtesy of Chris Fonnesbeck of the now defunct macinscience.org. My
scipy.test() still smokes badly, but now on other, fresh new issues. Apparently, I am not alone: getting SciPy to work on Mac seems to be something short of heroic.