Shared publicly  - 
ipython notebook

I just started working with ipython notebook a few days ago and already it is very useful for quick data processing. +Fernando Perez and +Brian Granger are working on it and making posts about it which I really like to see on google plus.


On my Ubuntu 11.04 box there is no libzmq1 just libzmq0, but the newer and required one is available in 11.10. I'm sure I could install it through a ppa but haven't bothered, I'll probably upgrade to 11.10 soon instead.

apt-get install libfreetype6-dev libpng-dev python-dev gfortran
I avoided the apt repositories where I thought easy_install would give me a newer version.

easy_install tornado scipy ipython matplotlib and numpy
Numpy produced a message "Parent module 'numpy.distutils' not found while handling absolute import" that I thought was an error but didn't impair me from using notebook.

After that I could ipython notebook --pylab inline and start working.


easy_install grabs the 0.10 scipy rather than the newer 0.11, but the documentation link within notebook is for 0.11. The one difference I've encountered so far is with griddata, which has 'nn' and 'linear' methods but not 'nearest' or 'cubic'. Also 'linear' in 0.10 is buggy, the gridded output comes out rotated. 'nn' also in some cases explodes producing values far outside the source data min/max. What is the easiest way to upgrade to scipy 0.11?

Text selection and navigation is in some cases unintuitive for me (at least in Chrome).

I wish ctrl-s would save the notebook, and that I could set a default so only the cleared notebook is saved. Running clear-all before saving is a must if the notebook is going into revision control, otherwise the notebook will have graphics and data inside it making much larger .pynb files that will generate enormous diffs upon committing them.

I like the inline plots, but is there a way to selectively make a regular interactive plot in that mode? Being able to do either would be great, I don't require the inline plots to be interactive themselves, just to be able to choose inline/not-inline-but-interactive for each plot.

I'm starting to see strange things in my data:
Fernando Perez's profile photoLucas Walter's profile photoConnor Johnson's profile photo
- Clear-before-saving: by all means file an issue in our tracker ( We've also toyed with the idea of splitting the nb in two files, one for inputs and one for large binary outputs, so that only the inputs are committed to VC. We'll have to experiment to see what works best, but version control is very much on our minds, and the clear-at-save option may be a good compromise.

- Interactive plotting: note that you can start without 'inline' and get normal interactive plots, which you can still 'paste' into the notebook by calling display(). I had a patch to allow toggling interactive/inline mode a while back but it's bit-rotted, I need to get back to that code to update it. There's also a project for an HTML5 canvas backend to matplotlib ( that preserves interactivity. We haven't done any integration with ipython yet, but this would be a great project to help with if you have some bandwidth!
Thanks- I've created an issue.

I'll start looking into the plotting software but contributing is a ways off.

One capability that really interests me (but perhaps understandibly very far from the aims of current developers) is more programmatically interactive plotting, allowing animations that can be altered live by executing cells, perhaps this would be done with cells with infinite for loops updating a plot with their data altered by other cells.
Do you know how to convert the printed window to a pdf? I haven't been able to figure anything out. I'm using Chromium with CrunchBang Linux. Thanks!
In Chrome and ipython 0.12 press the notebook print button which brings up a standalone browser window with about:blank in the url field that I can then right-click on it and select print (and then select pdf instead of a printer). The Chrome print screen comes up and does a reasonable formatting though plots are split across pages and not resized if they fall off the right edge, and the 'In [nn]:' takes up a lot of page width that could have helped fit the plots on the page.

(I haven't looked if issues exist for any of those problems yet)

Another annoying thing is the standalone chrome tab has no way to save as html (unless a certain key combination will do that?), I have to re-run the notebook in firefox to do that. See
Aha! That is a lot better! I have looked into a few other options for doing a similar task. I read that Opera lets you save pages as html. I'll keep you posted if I come up with anything. Thanks!
This is a really dirty and time-consuming option: I was able to save the image the way you had described, +Lucas Walter. I saved it as an .svg and then opened it up as multiple images in GIMP, at which point I cropped them onto an appropriately sized background, i.e., 850 pixels by 1100x( no. pages ) pixels. The crop tool blurred the edges a little bit, and the images had a white line running through them, but it turned out alright. Thanks for the advice!
Add a comment...