Skip to content

Plot disappear when blit is on #17685

Closed
Closed

Description

@Luluser

Already posted on stackoverflow but here should be more appropriate.

Bug report

Bug summary

First, good morning.
As the title describes, my plot disappears when the blitting option is on. I explain a bit further : I am making animations to display the solution of some differential equations and the code will become more and more heavy. I need the blitting option to have smooth animations
and not to "render" the widgets everytime but I also need a button to start/stop the animation. I am using FuncAnimation. The thing is when I stop the animation with the "myAnimation.event_source.stop()" command linked to a widget button, the plot disappears (in some cases the moment my mouse cursor gets out of the pause button, some cases as soon as I press the button). The ax stays empty as long as the animation is on pause and the plot comes back animated when I restart with "myAnimation.event_source.start()".
python_2020-06-19_10-04-11
Code for reproduction
A mock version of the code that is not the code I'm trying to write but simplifies the computations and makes it more readable.

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

class plotanimation:
    def __init__(self):
        self.fig,self.ax=plt.subplots()
        self.x=np.linspace(-10,10,1000)
        self.N=200
        self.interv=50
        self.n0=1./(4*np.pi*2e-4*0.1)**0.5 * np.exp(-self.x**2/(4*2e-4*0.1))  
        self.p,=self.ax.plot(self.x,self.n0)
        self.anim_running = True
        self.Myanimation=animation.FuncAnimation(self.fig, self.update,frames=self.N,interval=self.interv,blit=True)
    def update(self,i):
        self.n0+=i/100
        self.p.set_ydata(self.n0)
        return self.p,
    def animate(self):
        pause_ax = self.fig.add_axes((0.7, 0.025, 0.1, 0.04))
        pause_button = Button(pause_ax, 'pause', hovercolor='0.975')
        pause_button.on_clicked(self._pause)
        plt.show()
    def _pause(self, event):
        if self.anim_running:
            self.Myanimation.event_source.stop()
            self.anim_running = False
        else:
            self.Myanimation.event_source.start()
            self.anim_running = True
    
    
animated_plot = plotanimation()
animated_plot.animate()
# If applicable, paste the console output here
#
#

Expected outcome

The paused plot on the fig. Not an empty ax.

Matplotlib version

  • Operating system: W10
  • Matplotlib version: (3.1.2)
  • Matplotlib backend : TkAgg
  • Python version: v3.8.1:1b293b6
  • Jupyter version (if applicable): None
  • Other libraries: Numpy, scipy

Matplotlib was installed with pip. Conda is not installed.
Thank you a lot.

Activity

  1. tacaswell commented on Jun 19, 2020

    @tacaswell
    Member

    https://matplotlib.org/3.3.0/tutorials/advanced/blitting.html#sphx-glr-tutorials-advanced-blitting-py is probably of use here.

    The issue is that when using bliting, we mark the artists as "animated" via obj.set_animated(True) which means they are excluded from the normal draw process (so that you can get a "clean" background). In FuncAnimation we enable this (just to be on the safe side) to prevent artifacts where a previous data is "stuck" in the animation. When you pause the animation you do not un-set this state so when the figure redraws it skips rendering the plots (because it is the problem of what ever is managing the animation to do the render + blit).

    import numpy as np
    import matplotlib.pyplot as plt
    import matplotlib.animation as animation
    from matplotlib.widgets import Button
    
    
    class PlotAnimation:
        def __init__(self):
            self.fig, self.ax = plt.subplots()
            self.x = np.linspace(-10, 10, 1000)
            self.N = 200
            self.interv = 50
            self.n0 = (
                1.0
                / (4 * np.pi * 2e-4 * 0.1) ** 0.5
                * np.exp(-self.x ** 2 / (4 * 2e-4 * 0.1))
            )
            (self.p,) = self.ax.plot(self.x, self.n0)
            self.anim_running = True
            self.Myanimation = animation.FuncAnimation(
                self.fig, self.update, frames=self.N, interval=self.interv, blit=True
            )
    
        def update(self, i):
            self.n0 += i / 100 % 5
            self.p.set_ydata(self.n0 % 20)
            return (self.p,)
    
        def animate(self):
            pause_ax = self.fig.add_axes((0.7, 0.025, 0.1, 0.04))
            pause_button = Button(pause_ax, "pause", hovercolor="0.975")
            pause_button.on_clicked(self._pause)
            plt.show()
    
        def _pause(self, event):
            if self.anim_running:
                self.Myanimation.event_source.stop()
                self.p.set_animated(False)
                self.anim_running = False
                self.fig.canvas.draw_idle()
            else:
                self.p.set_animated(True)
                self.Myanimation.event_source.start()
                self.anim_running = True
    
    
    animated_plot = PlotAnimation()
    animated_plot.animate()

    works as expected.

    I really like this example! This is probably something that should be generalized and added to the Animation classes.

  2. Luluser commented on Jun 19, 2020

    @Luluser
    Author

    Thank you a lot for your solution, this is great help ! I had never found this page during my research, maybe you should attach a link on the doc page https://matplotlib.org/3.2.1/api/_as_gen/matplotlib.animation.FuncAnimation.html .

    Glad to read, feel free to use it !

    Also I have posted a logarithmic slider solution on stackoverflow (2nd answer to the question) : https://stackoverflow.com/questions/39960791/logarithmic-slider-with-matplotlib/62326375#62326375
    I think it would be great to have logarithmic sliders in matplotlib (as well as integer sliders, it's easy to make one but it often leads to rounding errors), I've used as much linear as log_10 sliders to be honest. I have feedback that my solution has a problem with the argument "orientation" that is not well recognized, maybe because I have not used an init function, I am not sure.
    Well again feel free to use it the way you want and to generalize to more log base if you like my recommendation.

    Thanks again.

  3. tacaswell commented on Jun 19, 2020

    @tacaswell
    Member

    @Luluser Can you open a PR with the text in the FuncAnimation docs that would have saved you and a PR adding the SliderLog?

  4. tacaswell commented on Jun 19, 2020

    @tacaswell
    Member

    and that link went live last night (that text was merged as part of the 3.3.0 release of which we just tagged RC1), it was just very good timing ;)

    I'm going to close this as resolved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions