OSX - "NumPy/SciPy requires Python 2.6 to Install"
Solution 1:
A simple and easy way to install numpy and scipy without going through version conflicts is to use the Enthought python distribution. They have a free version that includes python and the two packages you want, and the iPython console that I like a lot. They also offer a more extensive distribution that is free only for academic use, otherwise you have to pay for it.
python.org recommends installing an updated version of python 2.x instead of the version that comes with OSX. Following this advice, I was able to get scipy and numpy to work without enthought. Here are the versions that I am using (some version numbers missing, my notes are not complete):
- Python 2.7.2 from python.org (64 bit)
- numpy/scipy (OSX 10.6 build)
- ipython 0.12
- readline
- matplotlib
At the end of the day, enthought.com is the easiest way to install everything.
Solution 2:
Install latest version of Xcode/devtools and fortran ( http://www.scipy.org/Installing_SciPy/Mac_OS_X ).
Confirm python version
python --version
Install pip
sudo easy_install pip
Install numpy
sudo pip install numpy
If it complains you have an older version of bumpy installed, upgrade it.
pip install numpy --upgrade
check how many tests it fails ;)
sudo pip install nose
python
import numpy as np
np.test('full')
Solution 3:
If you have the OSX developer tools installed, you should be able to install from source without much trouble.
- make sure you have setuptools/distribute installed in your Python.org installation: run distribute_setup.py from the new python
- make sure you have the gfortran compiler for scipy. This is just
brew install gfortranif you use Homebrew. - use the right
easy_installto just doeasy_install numpy; easy_install scipy. (Or,easy_install pipand thenpip install numpy; pip install scipy.)