Matplotlib graphs do not clear in events with clear_output

Hi, I noticed this in 7.0.5, however it used to work in 6.0.0 with just clear_output, before it’s behaviour was changed. Below is a screenshot to replicate it; every time I click the button, the frame is cleared, but the graph is not. I would appreciate it if you can have a look.

Versions

jupyter                   1.0.0            py36h422fd7e_2
jupyter_client            5.1.0            py36h9902a9a_0
jupyter_console           5.2.0            py36h6d89b47_1
jupyter_core              4.3.0            py36h511e818_0
jupyterlab                0.27.0           py36h34cc53b_2
jupyterlab_launcher       0.4.0            py36h22c3ccf_0
jupyterthemes             0.18.2                    <pip>

ipywidgets                7.0.5                    py36_0    conda-forge
widgetsnbextension        3.0.8                    py36_0    conda-forge

matplotlib                2.1.0                     <pip>
pandas                    0.20.3           py36hce827b7_2

Screenshot

Code

import pandas as pd
import numpy as np
from IPython.display import clear_output
import matplotlib
from matplotlib import pyplot as plt
%matplotlib inline
import ipywidgets as widgets

button = widgets.Button(description="Plot!")
display(button)
out = widgets.Output()
display(out)

def on_button_clicked(b):
    with out:
        clear_output()
        x = pd.DataFrame(np.random.randn(8, 1))
        display(x.T)
        x.plot(figsize=[8, 1])
#         plt.plot(x) --- THIS ALSO CREATED DUPLICATES

button.on_click(on_button_clicked)

[jasongrout]

What’s happening here is that the matplotlib inline backend makes some assumptions about how to display graphs that don’t mesh well with capturing output. We had to work around this in interact as well at https://github.com/jupyter-widgets/ipywidgets/blob/02f406bd3ce1480ed3df2533f6acc1a4587b52af/ipywidgets/widgets/interaction.py#L75, which calls this function: https://github.com/jupyter-widgets/ipywidgets/blob/02f406bd3ce1480ed3df2533f6acc1a4587b52af/ipywidgets/widgets/interaction.py#L35

So you can use the same workaround we used. Import that function:

from ipywidgets.widgets.interaction import show_inline_matplotlib_plots

then change your function to:

def on_button_clicked(b):
    with out:
        clear_output()
        x = pd.DataFrame(np.random.randn(8, 1))
        display(x.T)
        x.plot(figsize=[8, 1])
        show_inline_matplotlib_plots()

Even better would be to use the matplotlib widget, which lets you update a matplotlib graph, respond to user input, etc: https://github.com/matplotlib/jupyter-matplotlib

[bybyte]

I was referring to the workaround not working for me. Yes, I understand – I will post in a new thread. Thanks.