<disclaimer: I work at msft>. It's been a pleasure working w the IPython team to get it working on Azure (engine on Linux or Windows VM) and Visual Studio. In case you want to check it out: https://www.windowsazure.com/en-us/develop/python/tutorials/...

Looking forward to trying out the updated notebook interface.

Any experience with it?

I tried to install and discuss the previous version with friends but failed. I couldn't find examples on the web, and installation with matplotlib wasn't working out of the box for Mac or Windows (not complaining, just observing it was not simple enough).

The IPython notebook's a real pleasure to use, and worth the trouble of getting it set up. The inline graphing works well, and if you take the trouble to give significant objects you care about a _repr_html_ method, you can use it as a very pleasant, visual data exploration tool.

I'll have to check it out. I didn't even know the IPython notebook existed until I saw this post, but it would've been incredibly useful in the past.

It was definitely a pain to get it running in the previous version on OSX, but entirely worthwhile. It's a really nice system that can replace mathematica and matlab for a lot of the work I do. In case this helps, here are the instructions that I ended up following to get the previous version running on OSX (using homebrew as my main package manager):

brew install zeromq

brew install qt

(add the following to .bash_profile):

export PYTHONPATH=/usr/local/lib/python:$PYTHONPATH

Download SIP 4.13 and PyQt4 from [1] and install with:

python configure.py

make

make install

Then do:

easy_install pygments

easy_install tornado

easy_install ipython[zmq,test]

easy_install --upgrade ipython

(from within ipython):

from IPython.external.mathjax import install_mathjax

install_mathjax()

[1] http://www.riverbankcomputing.co.uk/software/sip/download

Thank you!

It is absolutely amazing. I think it would be pure gold in the classroom.

With NotebookCloud, you can now run a fully featured IPython server on EC2 and set it up from your browser in a few clicks. This comes with support for inline plots, R, Octave, Cython, statsmodels and a ton of other features out the box. And it's free to use the service.

notebookcloud.appspot.com

The R magic is going to be a real win. Great job.

Agreed, but it's really a short-term win. Long term, python will need to (and I think it will) get better statistical libraries, that make R so great.

Statsmodels is a step in the right direction.