Hannes Bretschneider

If you haven’t been gripped by the IPython Notebook craze yet, let me quickly fill you in: IPython is a whole suite of tools whose goal it is to cover the whole scientific workflow from interactive data analysis to publication. It’s main focus, as is easy to guess, is Python, but they are ambitious about including other languages and already IPython can magically whizz data back and forth between Python, R, Octave, and soon, Julia.

Just this last Friday, after 11 years since the very first version of IPython and a breathtaking surge in popularity during the last year, the IPython team released version 1.0. This is a major milestone, mainly because of one feature, which many people might easily overlook: ipython nbconvert.

Let me explain: IPython Notebook is an interactive notebook that runs in your browser. It has a cell format, where each cell can either contain formatted text, or executable code. You can insert typeset mathematics, images, videos, arbitrary HTML, and pretty much everything you can imagine. Now, nbconvert will take a notebook and convert it to one of many output formats. So you can export to a static HTML page, a LaTeX document, Markdown, or even a slideshow running in your browser, enabled by Reveal.js. This may not sound like too much, but try to realize what this means: you now have an interactive programming environment that let’s you use the combined universes of all the Python, R, and Octave (Matlab) packages and use them for interactive data analysis. IPython Notebook keeps all your results, figures, and outputs in a single file, in a plain-text based format, which you can put under version-control, email, edit, and view on any platform. After your analysis is done you can run your notebook through a simple tool which will produce a publishable document in a myriad of formats. If you’re a scientist and you work with data, that’s just enormous.

Now one of the coolest new features are the Reveal.js based slideshows. Here is an example by the developer of the slideshow feature Damián Avila which shows how to turn any IPython Notebook into a slideshow and how to include math, images, videos, tables, etc. However, what’s slightly annoying about all the IPython Notebook formats, including the slideshow, is that there is no way to hide the code used to generate certain output. So, if you create a plot with matplotlib, your readers will always see the code you used to generate that figure. I don’t know why such a seemingly obvious feature hasn’t been implemented yet, but apparently there are some unspecified legal issues as hinted at by IPython developer Brian Granger.

In practice, your slides will look similar to this:

That’s great if you want to teach a class on Matplotlib, but quite often you’d want to hide that code cell.

Indeed, it is quite straight-forward to use a little bit of Javascript to simply hide the input cells in the browser when using the HTML or Reveal.js output. Here is a Javascript function (from Stack Overflow user Felix Kling) that hides any arbitrary class in a HTML document:

1
2
3
4
5
6
7

function hideElements(elements, start) {
    for(var i = 0, length = elements.length; i < length;i++) {
        if(i >= start) {
            elements[i].style.display = "none";
        }
    }
}

To hide the input cells we only need to hide the #input and optionally the #prompt classes. This is easily done by simply appending the following HTML to the output of nbconvert:

1
2
3
4
5
6
7
8
9
10
11
12
13

<script type="text/javascript">
function hideElements(elements, start) {
    for(var i = 0, length = elements.length; i < length;i++) {
        if(i >= start) {
            elements[i].style.display = "none";
        }
    }
}
var input_elements = document.getElementsByClassName('input');
hideElements(input_elements, 0);
var prompt_elements = document.getElementsByClassName('prompt');
hideElements(prompt_elements, 0);
</script>

Here’s the result on the slide from before:

Now, magically, you have a clean document, scrubbed of all code that you can go on to publish or present. Still, your code is still present in the document and the notebook, making your results always reproducable. The code is only hidden in the output.

To make this really simple, I created a small command line tool ipy_hide_input (download from gist). You can use the tool either in-place on a file, or on stdin like so:

1
2
3
4

ipy_hide_input input_file.slides.html # modify file in-place

# work on stdin
ipython nbconvert --to slids input.ipynb --stdout | ipy_hide_input > input.slides.html