Panel: Unable to redraw matplotlib plot when using interact

[ericmjl]

Unlike ipywidgets, the following minimally reproducible example does not appear to be working for me in the Jupyter notebook or Jupyter lab environment.

import matplotlib.pyplot as plt
import numpy as np

import panel as pn

from panel.interact import interact, interactive, fixed, interact_manual
from panel import widgets

pn.extension()

# from ipywidgets import interact

%load_ext autoreload
%autoreload 2
%matplotlib inline
%config InlineBackend.figure_format = 'retina'

def plot(minimum, maximum):
    x = np.arange(minimum, maximum)
    plt.plot(x, 3*x)


interact(plot, minimum=0, maximum=10)

The widgets display, the plot also displays on load, but scrubbing through the ranges does not update the plot.

Screenshot showing that the plot renders is below:

Given the similarity to the ipywidgets API, I was expecting this minimal example to work straight out of the box. I think I might be missing something conceptually here, but I am not quite sure. Would you be kind enough to enlighten me?

[jbednar]

Sure. Panel's interact function works with the return value of the given callback, which means that you have to ensure that whatever you want displayed is returned as the value, not displayed using whatever strange side effect %matplotlib inline uses to make the plot appear. I'm in no way an expert or even a casual Matplotlib user, so there is probably a much simpler way to get a proper Matplotlib figure out of a pyplot plotting call, but this should work:

import matplotlib.pyplot as plt
import numpy as np
import panel as pn
pn.extension()

def plot(minimum, maximum):
    x = np.arange(minimum, maximum)
    plot = plt.figure()
    plot.add_subplot(111).plot(x,3*x)
    plt.close(plot)
    return plot

pn.interact(plot, minimum=0, maximum=10)

[philippjfr]

I think you can get the current figure with plt.gcf()

[ericmjl]

@jbednar I see! Thank you for the pointers. That makes sense, thank you for correcting my misconception!

[jbednar]

Thanks, @philippjfr. That makes the answer much simpler:

import matplotlib.pyplot as plt
import numpy as np
import panel as pn
pn.extension()

def plot(minimum, maximum):
    x = np.arange(minimum, maximum)
    plt.plot(x,3*x)
    return plt.gcf()

pn.interact(plot, minimum=0, maximum=10)

I.e., add return plt.gcf() to the original code to return the current figure.

[philippjfr]

I did consider trying to replicate ipywidgets interact's behavior by looking for newly created matplotlib figures, but decided against it at the time.

[jbednar]

That would require spooky action at a distance, so I vote against it. I've added this question to the FAQ in https://github.com/pyviz/panel/pull/161. Thanks @ericmjl for bringing up the issue!