For my own reference. Brewed Python3, numpy, scipy and the rest goes in a pyvenv.

samuels_python3_setup.md

Samuel's Python3 setup for Mac:

This document describes in which way, what packages, and from where are installed on my Macs to get a decent Python 3.x, scipy/numpy & Co setup.

Note, I switched to Python's 3.x pyvenv from virtualenv in order to handle my pip-installable packages. Also thanks to Anaconda, the whole setup is so simple now that I use this document just to remember which additional packages I like.

---

• Install Homebrew

• Install the Anaconda Python distribution

• Optional Pyvenv (virtual env)

  • pyvenv --system-site-packages ~/Py to set up a virtual environment. Using the default symlink mode, has got an issue with brew's Python but I am using the executable from the Anaconda distribution now.
  • Don't forget to source ~/Py/bin/activate.
  • To upgrade when Anaconda has a newer Python: pyvenv --upgrade ~/Py

• Unit testing framework:

  • pip install coverage
  • pip install nose-progressive Nose plug-in with progress bars. Run with nosetests --with-progressive

• Progess bar

  • pip install pyprind

• Debugging - pip install debug

• OpenGL bindings:

  • pip install pyopengl

• VisPy: Most promising next gen visualizations

  • git clone https://github.com/vispy/vispy.git
  • cd vispy
  • python setup.py install

• GPU computing with numpy arrays:

  • git clone http://git.tiker.net/trees/pyopencl.git pyopencl
  • cd pyopencl
  • git submodule update --init
  • python configure.py --cl-enable-gl
  • python setup.py install

• Cython for writing C-extensions in Python-like speak

  • pip install cython

• Modular Data Processing Toolkit. (needs numpy):

  • git clone git://github.com/mdp-toolkit/mdp-toolkit
  • cd mdp-toolkit
  • python setup.py install

• Memory profiling and shell tools:

  • pip install psutil
  • pip install memory_profiler

• Read/write DICOM files (medical software) - pip install pydicom

  (or git clone and install by python3 setup.py install)

---

Todo

Install kivy

figure out best way to pass ars to pip so that the includes and libs of the homebrew stuff is found.

• PyYAML ... can't find libyaml. fuck.

• CFFI (C calling foregin functions. Needed for PyOpenCL if built from source)

  • brew install libffi
  • PKG_CONFIG_PATH=$(brew --prefix)/opt/libffi/lib/pkgconfig pip install cffi

• GUI toolkits:

  • todo: py3 ready?
  • brew install qt
  • brew install pyside or pyqt?
  • brew install vtk --python (the --qt option is currently not working on Xcode-only Macs)

• Audio and sound generation:

  • brew install liblo libsndfile portaudio portmidi
  • cd tmp
  • svn checkout http://pyo.googlecode.com/svn/trunk/ pyo
  • cd pyo
  • sudo python setup.py install --use-coreaudio --use-double
  • use scikit-audiolab instead?

ping @wonbyte :-)

I basically switched to Anaconda and therefore removed some of the manual install steps in this doc.

python in Anaconda seems built with difference options with system/homebrew versioned python. And I found it causing problem in packages where c-extensions are involved. Have you ever met similar problems?