Generating matplotlib graphs without a running X server

Matplotlib seems to require the $DISPLAY environment variable which means a running X server.
Some web hosting services do not allow a running X server session.
Is there a way to generate graphs using matplotlib without a running X server?

[username@hostname ~]$ python2.6
Python 2.6.5 (r265:79063, Nov 23 2010, 02:02:03)
[GCC 4.1.2 20080704 (Red Hat 4.1.2-48)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import matplotlib.pyplot as plt
>>> fig = plt.figure()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/username/lib/python2.6/matplotlib-1.0.1-py2.6-linux-i686.egg/matplotlib/pyplot.py", line 270, in figure
    **kwargs)
  File "/home/username/lib/python2.6/matplotlib-1.0.1-py2.6-linux-i686.egg/matplotlib/backends/backend_tkagg.py", line 80, in new_figure_manager
    window = Tk.Tk()
  File "/usr/local/lib/python2.6/lib-tk/Tkinter.py", line 1643, in __init__
    self.tk = _tkinter.create(screenName, baseName, className, interactive, wantobjects, useTk, sync, use)
_tkinter.TclError: no display name and no $DISPLAY environment variable
>>>

@Neil's answer is one (perfectly valid!) way of doing it, but you can also simply call matplotlib.use('Agg') before importing matplotlib.pyplot, and then continue as normal.

E.g.

import matplotlib as mpl
mpl.use('Agg')
import matplotlib.pyplot as plt

fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(range(10))
fig.savefig('temp.png')

You don't have to use the Agg backend, as well. The pdf, ps, svg, agg, cairo, and gdk backends can all be used without an X-server. However, only the Agg backend will be built by default (I think?), so there's a good chance that the other backends may not be enabled on your particular install.

Alternately, you can just set the backend parameter in your .matplotlibrc file to automatically have matplotlib.pyplot use the given renderer.

You need to use the matplotlib API directly rather than going through the pylab interface. There's a good example here:

http://www.dalkescientific.com/writings/diary/archive/2005/04/23/matplotlib_without_gui.html