Repository navigation
[ENH]: pickle (or save) matplotlib figure with insteractive slider #22482
Description
Problem
I'd like to save/pickle a matplotlib figure containing a matplotlib.widgets.Widget object to send it to a colleague without him having to run the code I use to generate it.
I know since matplotlib version 1.2 matplotlib.figure.Figure are pickable. However, when the figure contains a matplotlib.widgets.Widget, pickle fails to save it.
For example consider this class:
import matplotlib.pyplot as plt
import matplotlib.widgets as widgets
import numpy as np
import pickle as pkl
class interactive_plot():
def __init__(self, add_slider):
self.fig, self.axs = plt.subplots(2,1, gridspec_kw={'height_ratios':[10,1]})
self.x_plot = np.linspace(0,10,1000)
self.line = self.axs[0].plot(self.x_plot, self.x_plot*0)[0]
self.axs[0].set_ylim((0,10))
self.axs[0].grid()
self.axs[0].set_title(r'$y = x^\alpha$')
if add_slider:
self.slider= widgets.Slider(
self.axs[1],
label = r'$\alpha$',
valmin = 0,
valmax = 5,
valinit = 0,
valstep = .01
)
self.slider.on_changed(self.update)
def update(self, val):
self.line.set_ydata(self.x_plot**val)
If I run
h = interactive_plot(add_slider=False)
with open('test.pkl','wb') as f: pkl.dump(h, f)
all works properly. However
h = interactive_plot(add_slider=True)
with open('test.pkl','wb') as f: pkl.dump(h, f)
returns
Traceback (most recent call last):
File "C:\Users\user\Desktop\test.py", line 47, in <module>
with open('test.pkl','wb') as f: pkl.dump(h, f)
TypeError: cannot pickle 'FigureCanvasQTAgg' object
Is there a way to workaround this?
DISCLAIMER: The same question was posted here https://stackoverflow.com/questions/71145817/pickle-or-save-matplotlib-figure-with-insteractive-slider
Proposed solution
No response
Activity
anntzer commented on Feb 17, 2022
"fixes" the pickling itself (and seems like a reasonable patch regardless), but even then, the pickling process will drop any connected callbacks as functions can generally not be pickled.
mocquin commented on Jun 20, 2024
@anntzer is the dropping of callbacks when picking a figure builtin matplotlib ? or is it a limitation of the standard pickle module ? or is there some kind of criterion that make a function pickle-able or not ?
I am in the same kind of situation as @LucaAmerio, where I create matplotlib figure (with Qt backend) and callbacks connected to the figure and/or the axes. I'd like to be able to save and re-open those figure with the callbacks still active
anntzer commented on Jun 20, 2024
The ability to pickle functions is limited per https://docs.python.org/3/library/pickle.html#what-can-be-pickled-and-unpickled. We could consider pickling only whatever can actually be pickled and drop the rest, but I believe dropping everything (and asking whoever does the unpickling to reconnect things as needed) is not unreasonable either.