Plot update issue within event callback for multiple updates

[Apo-]

Platfrom/Tools Information:
Platform: Windows 10 x64
Python Version: 2.7.11
Python Distribution: Anaconda 2.5.0 (64-bit)
Matplotlib Version: 1.5.1

Issue Description:
When I use an event callback that updates a figure two (or more) times within the callback, only the last update on an element seems to be applied. Concretely, the plot is updated only when the event callback returns.

Example:
In the following example the expected outcome would be:

1. A figure with an axis with title "1"
2. When I press a key:
   --i. "Loading" is printed to the console and the axis title is updated to "Loading"
   --ii. After 3 seconds "2" is printed to the console and the axis title is updated to "2"

However, the observed outcome is:

1. A figure with an axis with title "1"
2. When I press a key:
   --i. "Loading" is printed to the console, but the axis title is not updated
   --ii. After 3 seconds "2" is printed to the console and the axis title is updated to "2"

from matplotlib import pyplot as plt
import time


def slow_plot(figure, axis):
    print "Loading"
    axis.set_title("Loading")
    figure.canvas.draw()

    time.sleep(3)

    print "2"
    axis.set_title("2")
    figure.canvas.draw()


def main():
    figure, axis = plt.subplots(1, 1)

    axis.set_title("1")
    figure.canvas.draw()
    figure.show()

    def on_key_press(event):
        slow_plot(figure, axis)

    figure.canvas.mpl_connect("key_press_event", on_key_press)

    plt.show()

if __name__ == "__main__":
    main()

Additional Information:

1. This code works as expected if I step through it step by step using the debugger.
2. Tried running the same piece of code on a Mac and I got the expected behavior.
3. I remember something similar working as expected about a year ago, so maybe this bug was introduced with matplotlib 1.5.1, but I can't say for sure. I'll try installing 1.4.* versions when I get some time and update the info.

[tacaswell]

The issue is integration with the GUI event loop. Remember that python is (basically) single threaded so if you are sleeping the GUI event loop is not spinning to re-paint your screen. On a mac I assume you are using the OSX backend which works slightly differently than the other backends which is why this works there. If you switch to QtAgg or TkAgg on the mac you will (most likely) get the same behavior.

You will not that during the sleep the figure is also 'inactive' in that you can not pan/zoom it (as all of the GUI events are piling up but not being processed). To re-draw the plot you just need to add a flush_events call (which spins the GUI event loop to clear out all of the events that have built up).

from matplotlib import pyplot as plt
import time

def slow_plot(figure, axis):
    print("Loading")
    axis.set_title("Loading")
    figure.canvas.draw()
    figure.canvas.flush_events()
    time.sleep(3)

    print("2")
    axis.set_title("2")
    figure.canvas.draw()



figure, axis = plt.subplots(1, 1)

axis.set_title("1")
figure.canvas.draw()
figure.show()

def on_key_press(event):
    slow_plot(figure, axis)

figure.canvas.mpl_connect("key_press_event", on_key_press)

plt.show()

Closing as I do not think there is a bug that we can do anything about in mpl.