Develop Reference

Running twice the libraries to have interactive plots on jupyter

Every time I want to have an interactive plot on Jupyter, I have to run the libraries twice. The libraries are as simple as this:
import matplotlib.pyplot as plt
%matplotlib notebook
The first time I make a plot I have a static image (like shown here). If I run again the libraries, the figure is now interactive. Do you know why?
Static image:
Interactive image:

You need to set the backend before importing pyplot. Therefore the order of your command matters. You will need to put any %matplotlib ... command before importing pyplot:
%matplotlib notebook
import matplotlib.pyplot as plt

matplotlib pyplot Invalid DISPLAY variable

While importing and attempting the following:
import matplotlib
from matplotlib import pyplot as plt
plt.plot([1,2,3],[1,4,9])
plt.show()
I get the following error. How do I fix? I am running Python 2.7, and notebook version 4.1.0. Thank you.
RuntimeError: Invalid DISPLAY variable

When running a jupyter notebook on a server, the server may not even be able to display the plot. The usual solution would be to use a non-interactive backend. In case of a jupyter notebook this would be done by adding
%matplotlib inline
at the top of the notebook, such that graphics are shown as png images.

Plot Window Hangs or is Empty in Python

I am having trouble with plots in a Jupyter Notebook in Python 3.5 on Mac OSX. The following code will hang when executed:
import numpy as np
import matplotlib.pyplot as plt
#%matplotlib inline
myfig = plt.plot(range(5))
plt.show()
If I restart the kernel and un-comment '%matplotlib inline', I do get plots to work inline. However, I'd like to be plotting in a separate window.
If I insert the following code at the beginning:
import matplotlib
matplotlib.use('Agg')
then restart the kernel and run, the code will not hang, but nothing will be plotted, no window opened.
Details:
Mac Book Pro running OSX El Capitan
Anaconda Python 3.5 in a Jupyter Notebook
backend is "MacOSX".

There is a post on GitHub mentioning that using Qt4Agg as backend worked...
If it is not available (and if you can), you might want to try using Hombrew to install Python (instead of Anaconda), Qt and/or Gtk with which you'll be able to use matplotlib without problem.

matplotlib crashing Python

I'm new to Python and matplotlib. A simple script I wrote is crashing and I was able to reproduce the crash with the following code:
import matplotlib.pyplot as plt
plt.figure(1)
plt.figure(2)
#plt.show()
The error is python.exe has stopped working. If I uncomment the plt.show(), it still crashes depending on the order I close the plots (no crash if 2 is closed first, crash if 1 is closed first). I'm using Windows 7, Python 3.4, and I installed the individual modules from www.lfd.uci.edu/~gohlke/pythonlibs/. Do I have something configured incorrectly or a misunderstanding of how to use matplotlib?

You need to set the TkAgg backend explicitly. With the following code, the problem is resolved.
import matplotlib
matplotlib.use("TkAgg")
from matplotlib import pyplot as plt
Note that setting the TkAgg backend after importing pyplot does not work either; it crashes too. You need to set it before importing pyplot.

I was having this issue, I thought it was some line in my code causing the bug, but in fact the very act of importing matplotlib.pyplot was killing my program. I solved it by first running it in verbose mode:
python -v [programname].py
This shows the last action the importer does before crashing. For me, the last line of this was:
import 'PyQt5' # <_frozen_importlib_external.SourceFileLoader object at 0x000001F8EC9C0908>
This suggested to me that the dependent library PyQt5 was causing issues, so I ran pip install PyQt5, and magically everything started working.

This could be issue with python 3.x
I have tried with python 2.7 on my windows machine and it works perfectly fine!
You can either downgrade your python to 2.7 or if you feel its too late to do why dont you give it a try to call close()
Import matplotlib
matplotlib.use('wxAgg')
Import matplotlib.pyplot as plt
# your scripts
plt.close('all')

I had a similar issue in OSX when I updated to Python 3.4. IDLE was also crashing and there was a warning telling me the version was unstable.
I solved it by following the prompts and updating the version of Tcl/Tk (8.5.9) - http://www.python.org/download/mac/tcltk .

For macOS, just make sure that
~/.matplotlib/matplotlibrc contains:
backend: MacOSX
You don't need the other backends unless you specifically want them. Alternatively, perhaps you can do:
import matplotlib
matplotlib.use("MacOSX")
though I have not tested that.

matplotlib does not work in Eclipse

I seem to have a problem that is in parts very similar to the one mentioned here:
Python with eclipse import problem
But unfortunatly just in parts otherwise that would have solved mine as well.
I use Eclipse SDK, Version: 3.7.0 with PyDev 101.
Furthermore I have installed
numpy-1.6.1rc1-win32-superpack-python2.6.exe
and
matplotlib-1.0.1.win32-py2.6.exe
as noted here:
http://matplotlib.sourceforge.net/users/installing.html
I have rebuild all the packages and looks the site-packages are listed.
(by the way as you see it is an Python version installed with ArcGIS )
If I test a script for instance a very simple one like:
import numpy
import matplotlib
import pylab as pl
I get the following error in Eclipse:
import matplotlib
import pylab as pl
from matplotlib.pylab import *
ImportError: No module named pylab
Even though the interpreter for Pydev is pointing to the appropriate version of python and matplotlib is installed properly in there (site-packages) it does not work in Eclipse. In iPython it works perfect.
What still needs to be done to get matplotlib work in Eclipse?
Thanks a lot!
Werner

pylab is in matplotlibs namespace, so this should work:
import matplotlib.pylab as pylab

I found that turning off interactive move and then calling show worked.
import matplotlib.pyplot as plt
#...your code...
plt.ioff()
plt.show()