Did you know that Anaconda has been supporting Python 3 for more than half a year? While the Anaconda installers use Python 2.7, it is quite easy to create new additional environments that provide Python 3.3 (and soon Python 3.4), as well as many scientific packages. This blog explains how this is done in detail.

When you install Anaconda, which is a versioned collection of useful big data and scientific packages, you have basically two options. The first is to use the full installer, which is several hundred megabytes, but includes everything in one download, and is therefore convenient to put on USB flash drives. The second is to use the mini-installer Miniconda, which installs only a minimal environment, and use the conda package manager to download and install subsequent packages as-needed.

Starting from a regular Anaconda installation

If you have installed Anaconda from the full installer, there is a root environment which has Python 2.7 and all other packages. What this means is that these packages are installed directly into the install prefix which was selected during the install process. The conda package manager now allows creating new environments separately from the root environment. Since these additional environments are completely separate, they may contain a different major Python version. The following command would be used to create an Anaconda Python 3 environment named py3k:

$ conda create -n py3k python=3 anaconda

Here python=3 and anaconda are package specifications, and it is the job of the SATsolver inside conda to find a consistent set of packages which satisfies these requirements. As the root environment uses Python 2, we had to specify the major version explicitly.

After adding the binary directory of the newly created environment to the PATH environment variable, which may be done using

$ source activate py3k

You are ready to use this Python 3 Anaconda environment.

Starting from Miniconda installation

Miniconda are installers which contain the conda package manager and Python. As conda works with either Python 2 or 3, Miniconda comes in two versions, i.e. the regular (which contains a Python 2 interpreter) and “Miniconda3” which comes with a Python 3 interpreter. This is the only difference between the two. So if you want a full Anaconda for Python 3 to be your root environment, you can first install Miniconda3, and then:

$ conda install anaconda

That is, you do not create a new environment, but simply install all Anaconda packages into your Python 3 root environment. You can still use the Miniconda3 setup to create Python 2 environments. However, as the default of the Python version is based on the Python version running conda, you now have to specify that you want to create an environment with Python 2 explicitly. For example:

$ conda create -n my_pandas python=2 pandas

Trying it out

Here is a small sample session, which shows how you create a Python 3 environment with IPython, activated it, and start it:

$ conda create -n py3k python=3 ipython

$ source activate py3k
prepending /home/ilan/a150/envs/py3k/bin to PATH
(py3k)$ ipython
Python 3.3.2 |Continuum Analytics, Inc.| (default, Aug  5 2013, 15:04:35) 
Type "copyright", "credits" or "license" for more information.

IPython 1.1.0 -- An enhanced Interactive Python.
?         -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help      -> Python's own help system.
object?   -> Details about 'object', use 'object??' for extra details.

In [1]:

So now that you’ve seen how easy it is to setup Python 3 with Anaconda, try creating a Python 3 environment yourself! Download Anaconda or Miniconda for free.