I gave a lightening talk in a recent Bay Area Python meet-up. I went over some of my recent hacks on combining ipython notebook and d3.js. What I wanted to show was how to mix python code and javascript code to create a dynamic programming/data analysis notebook. I created yet another example to demonstrate the great potential on combining the powerful tools.

If you are interested, you can try the this ipython notebook. You will need to download the development branch of the ipython v 0.13 to see the notebook. The notebook itself includes some of the explanation on how to run it and how it is done. I did not spend too much polishing the code and the motion chart, but it got the basic ingredients. If you want to peek it, here is a short screen recoding to show it looks like.

(click to download the movie)

Thanks for Brian Granger pointing out how to make bi-directional communication from javascript in a ipython-notebook front-end to back-end ipython kernel using the existing websocket/zmq channel architecture in ipython (see the thread ). I have been hacking around to see how to do it. I need to modified a few lines of the ipython-notebook javascript to make it work ( see my github commit ). I wrote some example to show how it works ( the ipython notebook and a screen shot ). Pretty cool that it works. It seems that one can develop a widget library to avoid hand-crafting both the javascript and python code for such communication. All right, one more small step toward to building some cool interactive visualization / analysis tools with ipython.

While I have been using ipython for a long time, I never really it more than just checking whether some code snippets working as expected. (Well, I tried to play with the parallel computing framework with ipython, but I never put it into production.) Just recently, I start to look into the ipython web-based notebook feature more carefully. It is great and make me think the ipython will make a python programmer or someone uses python for data analysis much more productive. (I used to envy the “RStudio” in the R-lang land, now, we python programmer finally have something more competitive.)

The cool thing using a web page as front-end is there are a lot potential using web interface for some cool visualization. I played with protovis.js a while ago. Recently, I went to a visualization meets-up, d3.js was mentioned a numbers of time. Then the idea comes to my mind “is it possible to combine the best of two world, python and d3.js?” After consulting some more experience users in the ipython-dev mailing list to see what is possible, I decided to spend some of my weekend time to hack it around. In the meantime, I get the chance to play with tornado, zero-mq and websocket, all the fun stuff these days. At the end, I am able to pass some javascript code written within the ipython notebook to get the browser to execute it and show some animation with d3.js. This will enable to create more fancier visualization in an interactive way all in a browser.

My weekend hacking results are hosted at github . I think there is a great potential to make thing like this working better. (For example, can we have a pythonic backend of d3.js? ) It definitely worth to mess it around to see more use like this.