##// END OF EJS Templates
making prompt_tolkit history search configurable
making prompt_tolkit history search configurable

File last commit:

r24087:c0f1d2be
r24145:61a08c53
Show More
Rich Output.ipynb
3310 lines | 293.5 KiB | text/plain | TextLexer

Rich Output

In Python, objects can declare their textual representation using the __repr__ method. IPython expands on this idea and allows objects to declare other, rich representations including:

  • HTML
  • JSON
  • PNG
  • JPEG
  • SVG
  • LaTeX

A single object can declare some or all of these representations; all are handled by IPython's display system. This Notebook shows how you can use this display system to incorporate a broad range of content into your Notebooks.

Basic display imports

The display function is a general purpose tool for displaying different representations of objects. Think of it as print for these rich representations.

In [1]:
from IPython.display import display

A few points:

  • Calling display on an object will send all possible representations to the Notebook.
  • These representations are stored in the Notebook document.
  • In general the Notebook will use the richest available representation.

If you want to display a particular representation, there are specific functions for that:

In [2]:
from IPython.display import (
    display_pretty, display_html, display_jpeg,
    display_png, display_json, display_latex, display_svg
)

Images

To work with images (JPEG, PNG) use the Image class.

In [3]:
from IPython.display import Image
In [4]:
i = Image(filename='../images/ipython_logo.png')

Returning an Image object from an expression will automatically display it:

In [5]:
i
Out[5]:
No description has been provided for this image

Or you can pass an object with a rich representation to display:

In [6]:
display(i)
No description has been provided for this image

An image can also be displayed from raw data or a URL.

In [7]:
Image(url='http://python.org/images/python-logo.gif')
Out[7]:
No description has been provided for this image

SVG images are also supported out of the box.

In [8]:
from IPython.display import SVG
SVG(filename='../images/python_logo.svg')
Out[8]:
No description has been provided for this image

Embedded vs non-embedded Images

By default, image data is embedded in the notebook document so that the images can be viewed offline. However it is also possible to tell the Image class to only store a link to the image. Let's see how this works using a webcam at Berkeley.

In [9]:
from IPython.display import Image
img_url = 'http://www.lawrencehallofscience.org/static/scienceview/scienceview.berkeley.edu/html/view/view_assets/images/newview.jpg'

# by default Image data are embedded
Embed      = Image(img_url)

# if kwarg `url` is given, the embedding is assumed to be false
SoftLinked = Image(url=img_url)

# In each case, embed can be specified explicitly with the `embed` kwarg
# ForceEmbed = Image(url=img_url, embed=True)

Here is the embedded version. Note that this image was pulled from the webcam when this code cell was originally run and stored in the Notebook. Unless we rerun this cell, this is not today's image.

In [10]:
Embed
Out[10]:
<IPython.core.display.Image object>

Here is today's image from same webcam at Berkeley, (refreshed every minutes, if you reload the notebook), visible only with an active internet connection, that should be different from the previous one. Notebooks saved with this kind of image will be smaller and always reflect the current version of the source, but the image won't display offline.

In [11]:
SoftLinked
Out[11]:
No description has been provided for this image

Of course, if you re-run this Notebook, the two images will be the same again.

HTML

Python objects can declare HTML representations that will be displayed in the Notebook. If you have some HTML you want to display, simply use the HTML class.

In [12]:
from IPython.display import HTML
In [13]:
s = """<table>
<tr>
<th>Header 1</th>
<th>Header 2</th>
</tr>
<tr>
<td>row 1, cell 1</td>
<td>row 1, cell 2</td>
</tr>
<tr>
<td>row 2, cell 1</td>
<td>row 2, cell 2</td>
</tr>
</table>"""
In [14]:
h = HTML(s)
In [15]:
display(h)
Header 1 Header 2
row 1, cell 1 row 1, cell 2
row 2, cell 1 row 2, cell 2

You can also use the %%html cell magic to accomplish the same thing.

In [16]:
%%html
<table>
<tr>
<th>Header 1</th>
<th>Header 2</th>
</tr>
<tr>
<td>row 1, cell 1</td>
<td>row 1, cell 2</td>
</tr>
<tr>
<td>row 2, cell 1</td>
<td>row 2, cell 2</td>
</tr>
</table>
Header 1 Header 2
row 1, cell 1 row 1, cell 2
row 2, cell 1 row 2, cell 2

JavaScript

The Notebook also enables objects to declare a JavaScript representation. At first, this may seem odd as output is inherently visual and JavaScript is a programming language. However, this opens the door for rich output that leverages the full power of JavaScript and associated libraries such as d3.js for output.

In [17]:
from IPython.display import Javascript

Pass a string of JavaScript source code to the JavaScript object and then display it.

In [18]:
js = Javascript('alert("hi")');
In [19]:
display(js)
SandBoxed(IPython.core.display.Javascript object)

The same thing can be accomplished using the %%javascript cell magic:

In [20]:
%%javascript

alert("hi");
SandBoxed(IPython.core.display.Javascript object)

Here is a more complicated example that loads d3.js from a CDN, uses the %%html magic to load CSS styles onto the page and then runs ones of the d3.js examples.

In [21]:
Javascript(
    """$.getScript('//cdnjs.cloudflare.com/ajax/libs/d3/3.2.2/d3.v3.min.js')"""
)
Out[21]:
SandBoxed(IPython.core.display.Javascript object)
In [22]:
%%html
<style type="text/css">

circle {
  fill: rgb(31, 119, 180);
  fill-opacity: .25;
  stroke: rgb(31, 119, 180);
  stroke-width: 1px;
}

.leaf circle {
  fill: #ff7f0e;
  fill-opacity: 1;
}

text {
  font: 10px sans-serif;
}

</style>
In [23]:
%%javascript

// element is the jQuery element we will append to
var e = element.get(0);
    
var diameter = 600,
    format = d3.format(",d");

var pack = d3.layout.pack()
    .size([diameter - 4, diameter - 4])
    .value(function(d) { return d.size; });

var svg = d3.select(e).append("svg")
    .attr("width", diameter)
    .attr("height", diameter)
  .append("g")
    .attr("transform", "translate(2,2)");

d3.json("data/flare.json", function(error, root) {
  var node = svg.datum(root).selectAll(".node")
      .data(pack.nodes)
    .enter().append("g")
      .attr("class", function(d) { return d.children ? "node" : "leaf node"; })
      .attr("transform", function(d) { return "translate(" + d.x + "," + d.y + ")"; });

  node.append("title")
      .text(function(d) { return d.name + (d.children ? "" : ": " + format(d.size)); });

  node.append("circle")
      .attr("r", function(d) { return d.r; });

  node.filter(function(d) { return !d.children; }).append("text")
      .attr("dy", ".3em")
      .style("text-anchor", "middle")
      .text(function(d) { return d.name.substring(0, d.r / 3); });
});

d3.select(self.frameElement).style("height", diameter + "px");
SandBoxed(IPython.core.display.Javascript object)

LaTeX

The IPython display system also has builtin support for the display of mathematical expressions typeset in LaTeX, which is rendered in the browser using MathJax.

You can pass raw LaTeX test as a string to the Math object:

In [24]:
from IPython.display import Math
Math(r'F(k) = \int_{-\infty}^{\infty} f(x) e^{2\pi i k} dx')
Out[24]:
$$F(k) = \int_{-\infty}^{\infty} f(x) e^{2\pi i k} dx$$

With the Latex class, you have to include the delimiters yourself. This allows you to use other LaTeX modes such as eqnarray:

In [25]:
from IPython.display import Latex
Latex(r"""\begin{eqnarray}
\nabla \times \vec{\mathbf{B}} -\, \frac1c\, \frac{\partial\vec{\mathbf{E}}}{\partial t} & = \frac{4\pi}{c}\vec{\mathbf{j}} \\
\nabla \cdot \vec{\mathbf{E}} & = 4 \pi \rho \\
\nabla \times \vec{\mathbf{E}}\, +\, \frac1c\, \frac{\partial\vec{\mathbf{B}}}{\partial t} & = \vec{\mathbf{0}} \\
\nabla \cdot \vec{\mathbf{B}} & = 0 
\end{eqnarray}""")
Out[25]:
\begin{eqnarray} \nabla \times \vec{\mathbf{B}} -\, \frac1c\, \frac{\partial\vec{\mathbf{E}}}{\partial t} & = \frac{4\pi}{c}\vec{\mathbf{j}} \\ \nabla \cdot \vec{\mathbf{E}} & = 4 \pi \rho \\ \nabla \times \vec{\mathbf{E}}\, +\, \frac1c\, \frac{\partial\vec{\mathbf{B}}}{\partial t} & = \vec{\mathbf{0}} \\ \nabla \cdot \vec{\mathbf{B}} & = 0 \end{eqnarray}

Or you can enter LaTeX directly with the %%latex cell magic:

In [26]:
%%latex
\begin{align}
\nabla \times \vec{\mathbf{B}} -\, \frac1c\, \frac{\partial\vec{\mathbf{E}}}{\partial t} & = \frac{4\pi}{c}\vec{\mathbf{j}} \\
\nabla \cdot \vec{\mathbf{E}} & = 4 \pi \rho \\
\nabla \times \vec{\mathbf{E}}\, +\, \frac1c\, \frac{\partial\vec{\mathbf{B}}}{\partial t} & = \vec{\mathbf{0}} \\
\nabla \cdot \vec{\mathbf{B}} & = 0
\end{align}
\begin{align} \nabla \times \vec{\mathbf{B}} -\, \frac1c\, \frac{\partial\vec{\mathbf{E}}}{\partial t} & = \frac{4\pi}{c}\vec{\mathbf{j}} \\ \nabla \cdot \vec{\mathbf{E}} & = 4 \pi \rho \\ \nabla \times \vec{\mathbf{E}}\, +\, \frac1c\, \frac{\partial\vec{\mathbf{B}}}{\partial t} & = \vec{\mathbf{0}} \\ \nabla \cdot \vec{\mathbf{B}} & = 0 \end{align}

Audio

IPython makes it easy to work with sounds interactively. The Audio display class allows you to create an audio control that is embedded in the Notebook. The interface is analogous to the interface of the Image display class. All audio formats supported by the browser can be used. Note that no single format is presently supported in all browsers.

In [27]:
from IPython.display import Audio
Audio(url="http://www.nch.com.au/acm/8k16bitpcm.wav")
Out[27]:

A NumPy array can be auralized automatically. The Audio class normalizes and encodes the data and embeds the resulting audio in the Notebook.

For instance, when two sine waves with almost the same frequency are superimposed a phenomena known as beats occur. This can be auralised as follows:

In [28]:
import numpy as np
max_time = 3
f1 = 220.0
f2 = 224.0
rate = 8000.0
L = 3
times = np.linspace(0,L,rate*L)
signal = np.sin(2*np.pi*f1*times) + np.sin(2*np.pi*f2*times)

Audio(data=signal, rate=rate)
Out[28]:

Video

More exotic objects can also be displayed, as long as their representation supports the IPython display protocol. For example, videos hosted externally on YouTube are easy to load:

In [29]:
from IPython.display import YouTubeVideo
YouTubeVideo('sjfsUzECqK0')
Out[29]:

Using the nascent video capabilities of modern browsers, you may also be able to display local videos. At the moment this doesn't work very well in all browsers, so it may or may not work for you; we will continue testing this and looking for ways to make it more robust.

The following cell loads a local file called animation.m4v, encodes the raw video as base64 for http transport, and uses the HTML5 video tag to load it. On Chrome 15 it works correctly, displaying a control bar at the bottom with a play/pause button and a location slider.

In [30]:
from IPython.display import HTML
from base64 import b64encode
video = open("../images/animation.m4v", "rb").read()
video_encoded = b64encode(video).decode('ascii')
video_tag = '<video controls alt="test" src="data:video/x-m4v;base64,{0}">'.format(video_encoded)
HTML(data=video_tag)
Out[30]:

External sites

You can even embed an entire page from another site in an iframe; for example this is today's Wikipedia page for mobile users:

In [31]:
from IPython.display import IFrame
IFrame('http://jupyter.org', width='100%', height=350)
Out[31]:

IPython provides builtin display classes for generating links to local files. Create a link to a single file using the FileLink object:

In [32]:
from IPython.display import FileLink, FileLinks
FileLink('Cell Magics.ipynb')

Alternatively, to generate links to all of the files in a directory, use the FileLinks object, passing '.' to indicate that we want links generated for the current working directory. Note that if there were other directories under the current directory, FileLinks would work in a recursive manner creating links to files in all sub-directories as well.

Rich output and security

The IPython Notebook allows arbitrary code execution in both the IPython kernel and in the browser, though HTML and JavaScript output. More importantly, because IPython has a JavaScript API for running code in the browser, HTML and JavaScript output can actually trigger code to be run in the kernel. This poses a significant security risk as it would allow IPython Notebooks to execute arbitrary code on your computers.

To protect against these risks, the IPython Notebook has a security model that specifies how dangerous output is handled. Here is a short summary:

  • When you run code in the Notebook, all rich output is displayed.
  • When you open a notebook, rich output is only displayed if it doesn't contain security vulberabilities, ...
  • ... or if you have trusted a notebook, all rich output will run upon opening it.

A full description of the IPython security model can be found on this page.

Rich output and nbviewer

Much of the power of the Notebook is that it enables users to share notebooks with each other using http://nbviewer.ipython.org, without installing IPython locally. As of IPython 2.0, notebooks rendered on nbviewer will display all output, including HTML and JavaScript. Furthermore, to provide a consistent JavaScript environment on the live Notebook and nbviewer, the following JavaScript libraries are loaded onto the nbviewer page, before the notebook and its output is displayed:

Libraries such as mpld3 use these capabilities to generate interactive visualizations that work on nbviewer.