mpld3: A D3 Viewer for Matplotlib

• Author: Jake Vanderplas jakevdp@cs.washington.edu
• License: BSD 3-clause

This is a minimal interactive viewer for matplotlib graphics built on D3. For an example of the code in action, see the blog post, or the IPython notebook examples available in the examples directory of this repository.

The code is still very incomplete, and does not (yet) support all of matplotlib's features. It should be considered a proof-of-concept, for now.

Installation

mpld3 requires jinja2 and matplotlib.

To install the library system-wide

[~]$ python setup.py install

Or, to install locally, use

[~]$ python setup.py install --prefix=/path/to/location/

Then make sure your Python path points to this location.

Trying it out

The package is pure python, and very light-weight. You can take a look at the notebooks in the examples directory, or run create_example.py, which will create a set of plots and launch a browser window showing interactive views of these plots.

For a more comprehensive set of examples, see the IPython notebook examples available in the examples directory.

Test Plots

To explore the comparison between D3 renderings and matplotlib renderings for various plot types, run the script process_testplots.py. This will generate an html page with the D3 renderings beside corresponding matplotlib renderings.

Features

Currently Supported

Currently the support of matplotlib features is very limited. The code supports the following:

• multiple axes, placed correctly on the figure
• lines and scatter plots created with plt.plot (not plt.scatter, etc.)
• grid lines and their properties
• title and axis labels
• patches (i.e. shapes like histograms, etc.)
• polygons (filled plots, etc.)
• some collections (scatter plots, etc.)

TODO List

There are many features still missing, and they range from fairly straightforward to fairly difficult.

• fine-grained text formatting (multi-lines, vertical alignment, etc.)
• containers (errorbars, etc.)
• axes aspect ratio
• tick specification & formatting
• twin axes (i.e. multiple scales on one plot) tied together
• adding a box-zoom tool
• adding tooltips

If any of these look like something you'd like to tackle, feel free to submit a pull request!