This walkthrough will cover most of what you’ll need to know in order to set up a functioning python installation, complete with ipython, and the scientific stack. We’ll look at the various libraries involved, such as numpy, pandas, and matplotlib, as well as one of the single greatest tools ever created: IPython.
Note: Currently there are a lot of great tutorials on using python, but I’ve found that most of the frustration involved with getting people up and running is setting up the environment for the first time. With that in mind, I’ve devoted
Contents
- Installing Homebrew
- Installing Python
- Changing System Paths
- Installing The Scientific Stack
- Tutorials
In order to get up and running, we’ll probably need Python. This guide will cover installation on OSX, but most of the homebrew commands can be replaced with equivalent package management tools for various “-nix” boxes. It isn’t quite as simple, but if you’re already developing on (for example) ubuntu or CentOS, then you get what you pay for.
All this aside, we’ll first need python, enter homebrew:
Installing Homebrew
$ ruby -e "$(curl -fsSL https://raw.github.com/mxcl/homebrew/go)"
...
$ brew update
...Homebrew is the “Missing Package Manager for OSX”. Some of you may have used MacPorts, or Fink… you have my sympathy.
Installing Python
The first rule of python is to never use the System version for anything. Homebrew is our friend, and it will be there beside us as we go down the rabbit hole. Worth mentioning: we’ll be using python 2.7.8 for this (if that offends you; you probably already know enough about python to forge off ahead on your own, I wish you all the best, friend).
$ brew install readline sqlite gdbm
...
$ brew install python --universal --frameworkThe --framework flag appends the version of python we’re using to the name, thus python becomes python2.7. the --universal flag tells homebrew (or possibly ‘make’) to build a universal binary.
Changing System Paths
$ cat /etc/paths
/usr/bin
/bin
/usr/sbin
/sbin
/usr/local/bin
$ vim /etc/paths
...In order for the homebrew libraries to take precedence over the system onces, we need to put /usr/local/bin (or whatever Homebrew prefix you’re using) at the top of that file.
This is important because as we start building/linking various libraries, it is possible that the system will start using the pre-existing libraries (like ancient versions of NumPy) instead of our Homebrewed ones.
You can do this in your favorite text editor.
You’ll also want to update your PATH.
Add these lines to your ~/.bash_profile.
The first makes sure you can see the Homebrew libraries, the second: makes it so virtualenv (if/when you use it) doesn’t fail silently.
export PATH=/usr/local/bin:$PATH
export PATH=/usr/local/share/python:$PATHMake sure you source ~/.bash_profile before you move on.
I personally run into a lot of problems installing a couple of the numeric libraries via pip. Installing scipy from source is nontrivial, but luckily Homebrew has our back …
Installing The Scientific Stack
In order to install these various packages, we need to source the appropriate homebrew cask:
$ brew tap homebrew/scienceThe Core libraries
These are the libraries that are absolutely, unequivocally required for any self-respecting numerical pythonista:
$ brew install numpy scipy matplotlibOptionally, if you’re into image processing, there are a couple other native libraries that are helpful:
$ brew install libpng libjpeg cairoInstalling Python libraries
IPython is the prompt that Python should have had, and the square-bracket-flag ([notebook]) tells pip to install the dependencies for that as well. The installation may or may not fail due to missing dependencies, but if you just ‘pip install _ “ the missing stuff until it works, you’ll be fine.
Scikit-learn is an absolutely fantastic, mature, and stable set of machine-learning packages.
I like to describe pandas as “R, for Python”, and it comes complete with efficient dataframe, and series datastructures, as well as glorious split-apply-combine semantics. Also: Wes McKinney is a really nice guy.
$ pip install ipython[notebook] scikit-learn pandas