Matplotlib

Using Matplotlib in Web Apps

I wanted to write a Python web app that would call Matplotlib to visualize some data on the back end, and serve it up to a browser window on the front end.

Initially I saw [webplotlib https://pypi.python.org/pypi/webplotlib/0.1], which looked promising, but wrapped all of matplotlib into two dinky kinds of plots: time series, and bar charts. I needed something that, like webplotlib, could communicate a figure to a browser, but something that, unlike webplotlib, still kept the full functionality of matplotlib.

The fix was easy. The core functionality of webplotlib is passing a figure as a string to the browser; this is about 4 lines. The rest is entirely case-dependent.

Let's walk through how to do this.

Step 1: Create Your Figure

Some quick code to make a dummy figure:

def make_dummy_figure():
    import matplotlib.pylab as plt
    from numpy.random import *

    fig = plt.figure()
    ax1 = fig.add_subplot(1,2,1)
    ax2 = fig.add_subplot(1,2,2)

    x = range(10)
    y1 = rand(10,)
    y2 = 1000*rand(10,)

    ax1.plot(x,y1,'b-')
    ax2.plot(x,y2,'r-')

    ax1.set_xlabel('Number of Llamas')
    ax1.set_ylabel('People killed')

    ax2.set_xlabel('Number of Tigers')
    ax2.set_ylabel('People killed')

Step 2: Make a Sendable Figure

To send a figure to our web application, we need to make the figure sendable. We modify the script to return a FigureCanvas handle to our figure,

import matplotlib.pylab as plt
from numpy.random import *

matplotlib.use('Agg')
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas

def make_dummy_figure():

    fig = plt.figure()
    ax1 = fig.add_subplot(1,2,1)
    ax2 = fig.add_subplot(1,2,2)

    x = range(10)
    y1 = rand(10,)
    y2 = 1000*rand(10,)

    ax1.plot(x,y1,'b-')
    ax2.plot(x,y2,'r-')

    ax1.set_xlabel('Number of Llamas')
    ax1.set_ylabel('People killed')

    ax2.set_xlabel('Number of Tigers')
    ax2.set_ylabel('People killed')

    return FigureCanvas(fig)

Step 3: Should've Put a String on it

Let's define a function that will turn that figure into a StringIO object:

def stringify_dummy_figure():
    figcanvas = make_dummy_figure()
    img_data_str = StringIO()
    figcanvas.print_png(img_data_str)
    img_data_str.seek(0)  # After writing, rewind data for further use.
    return img_data_str.read()

Step 4: Make an HTTP Response

The last step is to pass that string to an HTTP response

from django.http import HttpResponse, HttpResponseRedirect

img_str = stringify_dummy_figure()
response = HttpResponse(img_str, mimetype='image/png')

and that can be embedded into your web app, wherever it lays out the logic for parsing URLs and crafting HTTP responses.