Script to install Python scientific stack ("Scipy Superpack") using Homebrew and pip. Uses Homebrew's Python 2.7.6. Please report any issues in the comments.

install_superpack.sh

#!/bin/sh

hash brew &> /dev/null
if [ $? -eq 1 ]; then
    echo 'Installing Homebrew ...'
    ruby -e "$(curl -fsSL https://raw.github.com/Homebrew/homebrew/go/install)"
fi

# Ensure Homebrew formulae are updated
brew update

hash git &> /dev/null
if [ $? -eq 1 ]; then
    echo 'Installing Git ...'
    brew install git
fi

# Add science tap
brew tap homebrew/science

# Python tools and utilities
brew install python
brew install gfortran
pip install nose
pip install six
pip install patsy
pip install pygments
pip install sphinx
pip install cython

# IPython
brew install zeromq
pip install jinja2
pip install tornado
pip install pyzmq
pip install ipython

# OpenBLAS for NumPy/SciPy
brew install openblas
export BLAS=/usr/local/opt/openblas/lib/libopenblas.a
export LAPACK=/usr/local/opt/openblas/lib/libopenblas.a

# Build from cloned repo to avoid SciPy build issue
git clone [email protected]:numpy/numpy.git numpy_temp
cd numpy_temp
python setupegg.py bdist_egg
easy_install dist/*egg
cd ..
rm -rf numpy_temp

# SciPy
pip install git+git://github.com/scipy/scipy#egg=scipy-dev

# Matplotlib
brew install freetype
pip install git+git://github.com/matplotlib/matplotlib.git

# Rest of the stack
pip install git+git://github.com/pydata/pandas.git
pip install git+git://github.com/scikit-learn/scikit-learn.git
pip install git+git://github.com/pymc-devs/[email protected]
pip install git+git://github.com/statsmodels/statsmodels.git
pip install git+git://github.com/Theano/Theano.git

Very nice. Just a piece of unsolicited advice :)

Instead of tracking the master branch of all of these repositories, you should grab specific tags/commits. This reduces the chances that you'll end up with a bug in your stack that is impossible to track down because it could be coming from project X, Y, or Z.

Also, you should check out hashstack, we can handle a reproducible, versioned install profile like this in a similar amount of code (or less). I haven't put together an isolated SciPy stack, yet, but I'd be happy to do so if it would be useful.