Open source Enthought Python alternative [closed]
If you are one Windows, I can advise pythonxy for an easy and painless installation of Python and the core scientific libraries.
It is quite large and contains a lot of packages, which you maybe do not need, but at the installation, you can opt to choose which libraries to install.
You might at first exhale "what is he smoking?" to my answer, but here it comes as an echo to ogrisel's answer:
The best Python distribution is Debian GNU/Linux -- it comes with multiple versions of Python supported, hundreds (if not thousands) of Python modules and extensions packaged so their installation is guaranteed to be flawless (in 99% of the cases) regardless how complex underlying software/extension is, majority of them are unit-tested against supported versions and 3rd party modules at package build-time guaranteeing lack of head-ache later on.
Besides Python itself you can also choose there among a dozen of available Python IDEs (e.g. spyder, Eric, PIDA, and others), Python-aware editors (vim, emacs etc), alternative Python implementations (pypy), compilers (Cython, nuitka), etc. Debug build of Python (python-dbg) in tandem with gdb allow you right away debug your extensions while inspecting Python stack etc. And all of those Python-specific tools are available within the same software management framework as the rest of the system which carries thousands of generic and specialized software tools and resources.Depending on your demand you can choose between stable, testing and unstable Debian "suites".
Now "how": virtualization if you cannot or just prefer not to dual-boot. In a matter of minutes you can have a full blown system work on your Windows or OS X box without any major performance hit (unless you need heavy 3D graphics), and only need sufficient amount of RAM to share with your host OS needs -- you can easily access your host drive space within a virtual machine, see e.g. http://neuro.debian.net/vm.html for an easy starting point.
Ah right -- pandas, we provide backport builds from NeuroDebian repository, so you could easily use stable Debian and bleeding-edge pandas.