Developing on linux

Sketch of the steps to get up and running on linux to make binary releases for python packages. This is my personal setup.

My systems

  • Many. Currently I’m working on an Ubuntu Lucid 64 bit system:

    $ uname -a
    Linux angela 2.6.32-28-generic #55-Ubuntu SMP Mon Jan 10 23:42:43 UTC 2011 x86_64 GNU/Linux

Basic setup

sudo apt-get install git-core
sudo apt-get install python-dev python-numpy python-scipy
  • Install personal setup:

    git clone
    cd myconfig
    make dotfiles
    cd ..

    Then edit ~/.bashrc to add the commented lines at the top of your new ~/.bash_personal file. Then set up vim:

    git clone
    cd myvim
    sudo apt-get install ruby ruby-dev # for command-t
    make command-t
    make links

Setting up virtualenvs

  • Install virtualenv, and virtualenvwrapper. I did this with:

    sudo apt-get install python-setuptools
    sudo easy_install virtualenvwrapper

    I did this because there is was an incompatibility with Maverick Python 2.7 and Maverick virtualenv - see this bug report. For Natty (next after Maverick), you can probably use:

    sudo apt-get install virtualenvwrapper


  • If you are using my config (above), you probably want my default environment cleanup for virtualenvs:

    cd myconfig
    make virtualenvs
  • Make some good virtualenvs, with commands like:

    mkvirtualenv python25 --python=python2.5

    If you have to do something really bare, consider the no-site-packages flag to the mkvirtualenv command.

    You’ll probably want other versions of python for your virtualenvs. This:

    sudo apt-get install python2.7 python2.7-dev

    works on Ubuntu Maverick (10.10). You might want to use the old and new python versions repository to get e.g python 2.5 on new Ubuntus:

    sudo add-apt-repository ppa:fkrull/deadsnakes
    sudo apt-get update
    sudo apt-get install python2.5 python2.5-dev

    You may have to compile numpy and scipy from source for your non-system python versions:

    sudo apt-get build-dep python-numpy
    mkdir code
    cd code
    git clone git://
    cd numpy
    git co v1.5.1 # a tag
    sudo python2.5 install
    sudo rm -rf build
    sudo python2.7 install
    cd ..
    # This will soon become a real git repo rather than an svn mirror
    git clone --origin svn git:// scipy
    cd scipy
    git co svn/tags/v0.9.0rc4 # a tag
    sudo python2.5 install
    sudo rm -rf build
    sudo python2.7 install

    For the remaining packages, we can use setuptools. As per the instructions on old and new python versions

    sudo apt-get install python-setuptools-deadsnakes

    then (e.g.):

    sudo easy_install-2.5 nose