Fix plt.pause() behaviour

[thomasaarholt]

In #79, it is shown that plt.pause() raises an error when using ipympl.

%matplotlib widget
import matplotlib.pyplot as plt
fig = plt.figure()
plt.plot([1,2,3]) # this line isn't necessary
plt.pause(1)    
# AttributeError

In that issue, there is discussion recommending removing the two methods that raise the error, as the FigureCanvasBase.start_event_loop_default method does not exist. This PR implements that.

It turns out that the ipympl Canvas object actually inherits from FigureCanvasBase anyway (through Canvas -> FigureCanvasWebAggCore -> backend_agg.FigureCanvasAgg -> FigureCanvasBase), so removing them or renaming the called methods to remove the _default are equivalent.

Here is a gif showing the new behaviour with this PR (previous was an error)

[lgtm-com]

This pull request introduces 1 alert when merging 5079acc into f33046b - view on LGTM.com

new alerts:

• 1 for Unused import

[martinRenou]

Thanks a lot! Do you think it's also related to the other issue you opened (#290)?

There is a small flake8 issue F401 'matplotlib.backend_bases.FigureCanvasBase' imported but unused otherwise happy to merge it and make a patch release.

[thomasaarholt]

I removed that import now 👍 (Edit: I see that you merged - thanks!)

This PR is related to #290, but only in fixing the behaviour of plt.pause() so that it works like sleep() does.

Do you have any idea where the "blocking" behaviour in #290 sits? I was trying to think how qt does it. :/

[martinRenou]

Do you have any idea where the "blocking" behaviour in #291 sits? I was trying to think how qt does it. :/

Do you mean #290? Not sure, it looks related to other issues that are open. I think @ianhi looked at something similar concerning blocking comms or something. I suspect the comms are stuck and queued due to Matplolib, but I would need to look deeper into the code, I have to admit I never investigated this issue.

[thomasaarholt]

Sorry, yes #290. It's something that I think would be really nice - currently I'm tracking whether an algorithm is converging by generating plots as it runs, but I'd have to pre-generate all the figures for it to work with ipympl.

[mforbes]

@thomasaarholt In your example at the top, Cell #2 does not produce output., whereas when I run it now (matplotlib=3.10.1, ipympl=0.9.7) , the second cell produces 10 plots unless I use sleep(0.1) instead of plt.pause(0.1) as you have.