How do I force Matplotlib to draw while in the ipdb debugger in Spyder (or any other debugger)?

Question

EDIT

Unfortunately, at the moment this is not possible. I found out that it is a bug in Spyder. The developers are still figuring out how to approach this.

---

Goal

Visualize data while debugging code (and I want to use Spyder too!).

Attempt #1: Run foo.bar from IPython from Spyder

• Create a file named foo.py with the following code:

  from ipdb import set_trace as st
  import matplotlib.pyplot as plt
  
  def bar():
      st()
  

• While in IPython, type the following:

  In [4]: import foo
  
  In [5]: foo.bar()
  --Return--
  None
  > somewhere_over_the_rainbow\foo.py(5)bar()
        3 
        4 def bar():
  ----> 5     st()
  
  ipdb> plt.plot([1, 2], [3, 4])
  [<matplotlib.lines.Line2D object at 0x05CA8E90>]
  ipdb> plt.show()
  

Plot remains in "frozen" state. If I exit debugger, plot updates. If I try to close the plot, IPython crashes. Obviously both undesirable, and neither lets me see the data while debugging.

Attempt #2: Run foo.bar from IPython from command line

• Use same foo.py as in Attempt #1:

• Open IPython from commandline:

  In [4]: import foo
  
  In [5]: foo.bar()
  --Return--
  None
  > somewhere_over_the_rainbow\foo.py(5)bar()
        3
        4 def bar():
  ----> 5     st()
  
  ipdb> plt.plot([1, 2], [3, 4])
  [<matplotlib.lines.Line2D object at 0x03904070>]
  ipdb> plt.show()
  

Program shows plot as I expect. BUT I want to use Spyder.

Attempt #3: Run baz.bar from IPython from command line

• Write baz.py:

  from ipdb import set_trace as st
  import matplotlib.pyplot as plt
  
  st()
  

• Open IPython from commandline:

  In [4]: import baz
  --Return--
  None
  > somewhere_over_the_rainbow\baz.py(4)<module>()
        2 import matplotlib.pyplot as plt
        3 
  ----> 4 st()
  
  ipdb> plt.
  

Then Spyder fully freezes.

Any suggestions?

Note #1: In my full code, I have many files and many functions, so mashing it all together in one script without functions is not viable.

Note #2: Using any matplotlib interactive command (e.g. ion(), interactive(True), etc.) had no effect.

Note #3: Spyder version 2.0.12, Python 2.6, matplotlib 1.0.1.

Answer 1

(Spyder maintainer here) Spyder 4.2.0, released on November 8/2020, supports the ability to work with interactive Matplotlib plots while debugging. That works out of the box, i.e. it doesn't require setting any special option.

For previous versions, the best solution is to use the pause(n) command from Matplotlib (where n is a number of seconds) after showing the plot on ipdb. Here is an example:

from matplotlib.pyplot import imshow, pause
import numpy as np
x = np.random.rand(4,5)
imshow(x)
pause(1)

Answer 2

I found that you can actually plot in debug mode using Spyder now. It is surprisingly simple.

ipdb>pylab.plot(x,y)
ipdb>pylab.show()

......

Answer 3

Have you considered the ion() function when importing pylab? This should allow interactive plotting in pdb.

 import pylab
 import pdb
 pylab.ion()

 tst_xdata = [1,2,3,4,5,6]
 tst_ydata = [1,1,1,1,1,1]

 pylab.plot(tst_xdata,tst_ydata)
 pylab.draw()

 pdb.set_trace()
 for idx in range(3):

     tst_ydata = [elem+2 for elem in tst_ydata]
     pylab.plot(tst_xdata,tst_ydata)
     pylab.draw()

 pylab.show()

The above works on my machine (Ubuntu 11.04, Python 2.7, SciPy bersion 0.8.0), even running in Eclipse with PyDev.