Attempt to create a dynamically changing plot in console and notebook

[dsblank]

Bug report

Bug summary

Trying to have a dynamic plot that gets created and iteratively updated in a single function, both in the notebook, and in the console:

Python 3.6.1 |Anaconda custom (x86_64)| (default, May 11 2017, 13:04:09) 
Type 'copyright', 'credits' or 'license' for more information
IPython 6.1.0 -- An enhanced Interactive Python. Type '?' for help.

Other software versions:

• notebook (5.1.0)
• ipywidgets (7.0.0)
• jupyter (1.0.0)
• jupyter-client (5.1.0)
• jupyter-console (5.1.0)
• jupyter-core (4.3.0)

Code for reproduction

In the notebook, this is preceded with %matplotlib notebook. Testing on the Mac for the console part. All one cell.

import matplotlib.pyplot as plt
import numpy as np
import time

fig = plt.figure()
fig.show()

for j in range(10):            
    plt.plot(range(0, j), [1/(i+1) for i in range(0, j)])
    fig.canvas.draw()
    time.sleep(.05)      

Actual outcome

1. in the notebook, the figure size changes (initial is small, final fills area)
2. in the console, the plot doesn't show up until all done plotting

Matplotlib version

• Operating system: MacOSX, Darwin 16.7
• Matplotlib version: matplotlib (2.0.2)
• Matplotlib backend: MacOSX
• Python version: 3.6
• Jupyter version (if applicable): 1.0.0
• Other libraries: (see above), anaconda install

[jhamrick]

in the notebook, the figure size changes (initial is small, final fills area)

I just ran into this bug as well. During the loop, the figure is small, e.g.:

when the loop finishes, the figure fills the whole area, e.g.:

This was not a problem in matplotlib 1.5.3 (i.e. the plot always filled the whole area), and seems to have been introduced with the release of 2.0.0.

[tacaswell]

The issue is resizing is the high-dpi support and a race condition of when the front end JS resizes to account for it and when the backend gets notified of it.

[dsblank]

Note that is just one of the issues; the other is that it doesn't show in the console at all, until the code is finished.

[tacaswell]

If you add a show() it should work in the console and I don't think it will break in the notebook.

[dsblank]

Adding fig.show() does allow it to work on my Linux console, but not on the MacOSX backend.

However, adding fig.show() works on the MacOSX backend, but not in the notebook because:

1. this issue: plt.pause() with notebook backend causes error  #9729
2. it would probably still have the "race condition" size problem

Maybe I'm going about this all wrong? Perhaps there is a more standard method of updating a plot on all GUI backends?

[dsblank]

One problem left: the tiny version, followed by the updated, full-frame version.

I don't believe that this is a "race condition" but rather a design issue. It appears that fig.show() is sending information to the backend, but (if done all in one cell) the program is running, and can't respond to it.

For example, if a time.sleep() is placed after fig.show() in this updated example:

import matplotlib.pyplot as plt

fig = plt.figure()
fig.show()
sleep.time(5) ### added time here!
for j in range(10):            
    plt.plot(range(0, j), [1/(i+1) for i in range(0, j)])
    plt.pause(.05)

The small version shows, followed by the full-size version as @jhamrick showed above. However, if split into two cells (at the inserted time), then t works as desired (size is correct).

I don't know how to debug this one.

[lucainnocenti]

I think this is the same as matplotlib/ipympl#17

[streamnsight]

if found this that worked for me.
https://stackoverflow.com/questions/21360361/how-to-dynamically-update-a-plot-in-a-loop-in-ipython-notebook-within-one-cell