matplotlib does not show my drawings although I call pyplot.show()

Solution 1:

If I set my backend to template in ~/.matplotlib/matplotlibrc, then I can reproduce your symptoms:

~/.matplotlib/matplotlibrc:

# backend      : GtkAgg
backend      : template

Note that the file matplotlibrc may not be in directory ~/.matplotlib/. In this case, the following code shows where it is:

>>> import matplotlib
>>> matplotlib.matplotlib_fname()

In [1]: import matplotlib.pyplot as p

In [2]: p.plot(range(20),range(20))
Out[2]: [<matplotlib.lines.Line2D object at 0xa64932c>]

In [3]: p.show()

If you edit ~/.matplotlib/matplotlibrc and change the backend to something like GtkAgg, you should see a plot. You can list all the backends available on your machine with

import matplotlib.rcsetup as rcsetup
print(rcsetup.all_backends)

It should return a list like:

['GTK', 'GTKAgg', 'GTKCairo', 'FltkAgg', 'MacOSX', 'QtAgg', 'Qt4Agg',
'TkAgg', 'WX', 'WXAgg', 'CocoaAgg', 'agg', 'cairo', 'emf', 'gdk', 'pdf',
'ps', 'svg', 'template']

Reference:

• Customizing matplotlib

Solution 2:

I ran into the exact same problem on Ubuntu 12.04, because I installed matplotlib (within a virtualenv) using

pip install matplotlib

To make long story short, my advice is: don't try to install matplotlib using pip or by hand; let a real package manager (e.g. apt-get / synaptic) install it and all its dependencies for you.

Unfortunately, matplotlib's backends (alternative methods for actually rendering your plots) have all sorts of dependencies that pip will not deal with. Even worse, it fails silently; that is, pip install matplotlib appears to install matplotlib successfully. But when you try to use it (e.g. pyplot.show()), no plot window will appear. I tried all the different backends that people on the web suggest (Qt4Agg, GTK, etc.), and they all failed (i.e. when I tried to import matplotlib.pyplot, I get ImportError because it's trying to import some dependency that's missing). I then researched how to install those dependencies, but it just made me want to give up using pip (within virtualenv) as a viable installation solution for any package that has non-Python package dependencies.

The whole experience sent me crawling back to apt-get / synaptic (i.e. the Ubuntu package manager) to install software like matplotlib. That worked perfectly. Of course, that means you can only install into your system directories, no virtualenv goodness, and you are stuck with the versions that Ubuntu distributes, which may be way behind the current version...