Is there any test that can be added in mpl (maybe a mock of some sort) for this?

working on a minimal example to reproduce atm

It is not a typo, both button and buttons are valid.

Still trying to wrap my head around a minimal working example, getting differing behaviour in a reduced case vs my application...

But if button and buttons are both valid, and one works, any reason not to change it?

But if button and buttons are both valid, and one works, any reason not to change it?

They do different things. See #28453.

Still working at an example, it's pretty finicky to reproduce so makes me wonder if there is something else going on under the hood, but forcing an error output in that block of code yields:

287 button = event['button'] + 1  # JS numbers off by 1 compared to mpl.
    288 print("event:")
--> [289] raise Exception(repr(event))
    290 buttons = {  # JS ordering different compared to mpl.
    291     button for button, mask in [
    292         (MouseButton.LEFT, 1),
   (...)    297     ] if event['buttons'] & mask  # State *after* press/release.
    298 }
    299 modifiers = event['modifiers']

Exception: {'x': 441, 'y': 185, 'button': 0, 'step': 1, 'modifiers': [], 'guiEvent': {'isTrusted': True, 'data': 'scroll', 'step': 1}, 'type': 'scroll'}

Which at least shows for some reason 'buttons' is not in the event.

I'm happy to move this over to an issue for more discussion on ipympl or here, but I believe there is something still not right.

I'm happy to move this over to an issue for more discussion on ipympl or here, but I believe there is something still not right.

There is something wrong in the sense that this is fixed in ipympl at these lines
https://github.com/matplotlib/ipympl/blob/5e068ed222427919cc20c406e34d02ba9be55cc3/src/mpl_widget.ts#L760-L761

but some people using ipympl 0.9.7 are still seeing problems.

Personally I am not able to reproduce the problem. I use ipympl and matplotlib in jupyterlab frequently, using all 3 major OSses and different versions of python, and installing from PyPI and conda-forge. To make progress what is needed is a reliable reproducer, i.e. the steps needed to get a system into the state that ipympl 0.9.7 does not work.

Minimal expample on my machine:

import matplotlib
matplotlib.use('module://ipympl.backend_nbagg')  # Set the backend programmatically

import matplotlib.pyplot as plt
import numpy as np

# Data
x = np.linspace(0, 2 * np.pi, 100)
y = np.sin(x)

fig, ax = plt.subplots()

ax.plot(x, y, label='Sine Wave')

plt.show()

I'm looking to interactively view on vscode server, via remote ssh. I've blown away .vscode-server to ensure a clean environment server side, with a fresh venv of python 3.12.11 pip freeze yields:

asttokens==3.0.0
comm==0.2.2
contourpy==1.3.2
cycler==0.12.1
debugpy==1.8.15
decorator==5.2.1
executing==2.2.0
fonttools==4.59.0
ipykernel==6.30.0
ipympl==0.9.7
ipython==9.4.0
ipython_pygments_lexers==1.1.1
ipywidgets==8.1.7
jedi==0.19.2
jupyter_client==8.6.3
jupyter_core==5.8.1
jupyterlab_widgets==3.0.15
kiwisolver==1.4.8
matplotlib==3.10.3
matplotlib-inline==0.1.7
nest-asyncio==1.6.0
numpy==2.3.1
packaging==25.0
parso==0.8.4
pexpect==4.9.0
pillow==11.3.0
platformdirs==4.3.8
prompt_toolkit==3.0.51
psutil==7.0.0
ptyprocess==0.7.0
pure_eval==0.2.3
Pygments==2.19.2
pyparsing==3.2.3
python-dateutil==2.9.0.post0
pyzmq==27.0.0
six==1.17.0
stack-data==0.6.3
tornado==6.5.1
traitlets==5.14.3
wcwidth==0.2.13
widgetsnbextension==4.0.14

Hope this might help?

Full steps for me to reproduce reliably:

1. mv/ delete .vscode-server directory
2. check ps aux | grep jupyter and kill any lingering <pid>
3. fully close down vscode
4. killall -u <username> on remote machine, forcing a disconnect
5. ssh connect to remote, create test directory, cd in
6. python3 -m venv testEnv
7. source testEnv/bin/activate
8. pip install ipykernel matplotlib ipympl numpy
9. Connect to ssh host on vscode
10. Open test directory in vscode
11. Create test.ipynb with above code, and save
12. Select kernel -> install recommended extensions [jupyter + python]
13. Reload window
14. Select kernel -> testEnv
15. Run all

Returns:

286 e_type = event['type']
    287 button = event['button'] + 1  # JS numbers off by 1 compared to mpl.
    288 buttons = {  # JS ordering different compared to mpl.
    289     button for button, mask in [
    290         (MouseButton.LEFT, 1),
    291         (MouseButton.RIGHT, 2),
    292         (MouseButton.MIDDLE, 4),
    293         (MouseButton.BACK, 8),
    294         (MouseButton.FORWARD, 16),
--> [295] if event['buttons'] & mask  # State *after* press/release.
    296 }
    297 modifiers = event['modifiers']
    298 guiEvent = event.get('guiEvent')

Can you try using JupyterLab instead of VSCode?

That doesn't throw any errors, so could be in the generation of the javascript sent to vscode?

It seems likely that vscode is the key factor here, given my observation that I've never experienced a problem and I always use JupyterLab rather than vscode.

I don't know what vscode does in these situations. My understanding is that it (and others such as PyCharm) reuse the Jupyter machinery but modify it for their own purposes. So it is quite possible that whatever they do has not been upgraded to work with ipympl 0.9.7 correctly.

I close this as it seems there is nothing we can do on the matplotlib side, and the PR itself is certainly not the correct solution.