Jupyter Notebook matplotlib notebook makes plot not show up, inline makes it not interactive

Question

I have one python file Vis.py with the following two functions:

import matplotlib.pyplot as plt
from matplotlib.widgets import Slider
def update(val): #needed for slider function of plot_test
 pos = spos.val
 ax.axis([pos,pos+10,-1,1])
 fig.canvas.draw_idle()
def plot_test(data):
 fig, ax = plt.subplots()
 plt.subplots_adjust(bottom=0.25)
 plt.plot(data)
 plt.axis([0, 10, -1, 1])
 axcolor = 'lightgoldenrodyellow'
 axpos = plt.axes([0.2, 0.1, 0.65, 0.03], facecolor=axcolor)
 spos = Slider(axpos, 'Pos', 0.1, 90.0)
 spos.on_changed(update)
 plt.show();

and I am trying to use the plot_test function in a separate ipynb file:

%matplotlib notebook
from Vis import *
import numpy as np
t = np.arange(0.0, 200.0, 0.1)
s = np.sin(2*np.pi*t)
plot_test(s)

However, the plot doesn't show up, not even an empty white space. I tried running %matplotlib inline before plot_test(s). That makes the plot show up, but it also gets rid of the interactiveness of the plot.

• ImportanceOfBeingErnest: It works fine for me; though to get an interactive slider you would need to define ax in the updating function. Best make update part of plot_test(), such that ax is accessible in the namespace.
• emilaz: Huh strange. So you can use the slider on the bottom, using my code as-is? Regarding the update function, I'd agree with you, however the update function is being used as in my example all over the internet (matplotlib.org/gallery/widgets/slider_demo.html, stackoverflow.com/questions/31001713/…). If you can provide an explicit example with the update function being part of the plot_test() function and working interactively, I'll gladly accept it as an answer.

Answer 1

The updating function references ax, which is out of scope. A solution is to put the updating function inside the plot_test function.

import matplotlib.pyplot as plt
from matplotlib.widgets import Slider
def plot_test(data):
 fig, ax = plt.subplots()
 plt.subplots_adjust(bottom=0.25)
 plt.plot(data)
 plt.axis([0, 10, -1, 1])
 axcolor = 'lightgoldenrodyellow'
 axpos = plt.axes([0.2, 0.1, 0.65, 0.03], facecolor=axcolor)
 spos = Slider(axpos, 'Pos', 0.1, 90.0)
 def update(val): #needed for slider function of plot_test
 pos = spos.val
 ax.axis([pos,pos+10,-1,1])
 fig.canvas.draw_idle()
 spos.on_changed(update)
 plt.show()

Then, keeping the notebook part unchanged,

%matplotlib notebook
from Vis import *
import numpy as np
t = np.arange(0.0, 200.0, 0.1)
s = np.sin(2*np.pi*t)
plot_test(s)

results in the desired interactive figure for me.

enter image description here

• emilaz: Perfect, thanks! A small thing to add: For me, the plot still didn't show up, but adding %matplotlib notebook a second time after the from Vis import * line weirdly solved it for me. Might help someone in the future.
• ImportanceOfBeingErnest: That's when you don't start with a fresh kernel, i.e. if have used the inline backend before, changed the line to notebook, but did not restart the kernel. See this answer.
• emilaz: No that wasn't it. I got rid of the inline statement and restarted the kernel, but it's not showing without the second %matplotlib notebook statement. Also, in the question you referenced, a blank white space appears, which for me wasn't the case.
• ImportanceOfBeingErnest: Your notebook may be configured in a way to automatically load the inline backend. This seems to often be the case (though I don't know why that is).