search

matplotlib / ipympl

Matplotlib Jupyter Integration
https://matplotlib.org/ipympl/
BSD 3-Clause "New" or "Revised" License
1.59k stars 225 forks source link

ipympl 0.8.2 - ginput() does not work on Jupyterlab / returns an empty array #400

Open opens9999 opened 2 years ago

opens9999 commented 2 years ago

Describe the bug I should be able to select data points from a figure and then red crosses are supposed to appear after clicking with the mouse, but nothing happens, an empty array is returned and the figures only appear after the timeout of ginput().

Expected behavior Figure pops up and I can click with the mouse on the figure to select data points. An array will be generated

To Reproduce

%matplotlib widget 

import matplotlib.pyplot as plt
import numpy as np
t = np.arange(10)

plt.plot(t, np.sin(t))

plt.title('matplotlib.pyplot.ginput()\
 function Example', fontweight ="bold")

print("After 3 clicks :")
x = plt.ginput(3)
print(x)

plt.show()

My Platform

Firefox on Arch Linux, latest version after system update. I tested it on Jupyterlab, I did not test it on Jupyter-notebook.

ianhi commented 2 years ago

This runs into a current limitations of ipywidgets (which ipympl is based on) which is that blocking input is impossible. Until that is fixed then ginput will not behave as expected for either ipympl or nbagg.

We should probably at least raise a warning about this but I don't think there's really a safe reasonable way for a backend to reach into pyplot and mess with things (@tacaswell ?)

As a workaround I suggest building your code around the matplotlib callback system, in particular to make use of the pick event. For example: https://matplotlib.org/stable/gallery/event_handling/pick_event_demo.html

Or see how scatter_selector works https://github.com/ianhi/mpl-interactions/blob/3d98c9b7f603d12ae082dd2e90c3ee6ae943f874/mpl_interactions/widgets.py#L48