BUG: Unexpected plot discrepancy when plotting as script compared to running in IPython

[Gnefil]

Description of the problem

Discrepancies between running figure plots as a Python script instead of running in IPython.
See fig1, fig2, fig3 as examples below.
This has been tested in a minimal plotting example including mne.viz.plot_topomap, stc.plot, mne.viz.plot_evoked_topo. In one case, topomap and stc plot are not synchronised when they should, in another case, the stc plot wouldn't even show up. All these happening when run as script, but works fine when run in IPython.

Steps to reproduce

# Working example when fig1 and fig2 plots. Synchronized. Both figures appear.

import matplotlib.pyplot as plt
import mne
data_path = mne.datasets.sample.data_path()

evoked_fname = data_path / "MEG/sample/sample_audvis-ave.fif"
evoked = mne.read_evokeds(evoked_fname, condition="Left Auditory").apply_baseline()

stc_fname = data_path / "MEG" / "sample" / "sample_audvis-meg-eeg"
stc = mne.read_source_estimate(stc_fname)


fig1 = evoked.plot_topomap("interactive")
fig2 = stc.plot("sample", subjects_dir=data_path / "subjects")
# fig3 = mne.viz.plot_evoked_topo(evoked)
link(fig1, fig2)

plt.show()


# Broken example, when run as a Python script. fig1 and fig2 no longer synchronised. All three figures appear.

import matplotlib.pyplot as plt
import mne
data_path = mne.datasets.sample.data_path()

evoked_fname = data_path / "MEG/sample/sample_audvis-ave.fif"
evoked = mne.read_evokeds(evoked_fname, condition="Left Auditory").apply_baseline()

stc_fname = data_path / "MEG" / "sample" / "sample_audvis-meg-eeg"
stc = mne.read_source_estimate(stc_fname)


fig1 = evoked.plot_topomap("interactive")
fig2 = stc.plot("sample", subjects_dir=data_path / "subjects")
fig3 = mne.viz.plot_evoked_topo(evoked)
link(fig1, fig2)


plt.show()


# Broken example, when run as a Python script. fig1 and fig2 no longer synchronised. Only fig1 and fig3 appear; fig2 won't appear.

import matplotlib.pyplot as plt
import mne
data_path = mne.datasets.sample.data_path()

evoked_fname = data_path / "MEG/sample/sample_audvis-ave.fif"
evoked = mne.read_evokeds(evoked_fname, condition="Left Auditory").apply_baseline()

stc_fname = data_path / "MEG" / "sample" / "sample_audvis-meg-eeg"
stc = mne.read_source_estimate(stc_fname)


fig1 = evoked.plot_topomap("interactive")
fig3 = mne.viz.plot_evoked_topo(evoked)
fig2 = stc.plot("sample", subjects_dir=data_path / "subjects")
link(fig1, fig2)


plt.show()


# Broken examples are run with `py ./location_to_script.py`. But when run with IPython, i.e. `ipython` command, followed by `%matplotlib qt` and `%run ./location_to_script.py`, the examples are not broken.

Link to data

No response

Expected results

The plots should behave properly, showing up and being synchronised even with fig3, just like running in IPython.

Actual results

Unexpected behaviour, as described in examples.

Additional information

Platform Windows-11-10.0.26200-SP0
Python 3.12.7 (tags/v3.12.7:0b05ead, Oct 1 2024, 03:06:41) [MSC v.1941 64 bit (AMD64)]
Executable D:\cloud\Projects\mne-python.venv\Scripts\python.exe
CPU Intel(R) Core(TM) i9-14900HX (32 cores)
Memory 63.6 GiB

Core

• mne 1.12.0.dev128+gcdbdf9200 (unable to check for latest version on GitHub, timeout after 2.0 sec)
• numpy 2.4.3 (OpenBLAS 0.3.31.dev with 24 threads)
• scipy 1.17.1
• matplotlib 3.10.8 (backend=qtagg)

Numerical (optional)

• sklearn 1.8.0
• nibabel 5.4.2
• dipy 1.12.0
• pandas 3.0.1
• h5py 3.16.0

• unavailable numba, nilearn, openmeeg, cupy, h5io

Visualization (optional)

• pyvista 0.47.1 (OpenGL 4.5.0 - Build 32.0.101.6790 via Intel(R) RaptorLake-S Mobile Graphics Controller)
• pyvistaqt 0.11.3
• vtk 9.6.1
• qtpy 2.4.3 (PySide6=6.11.0)

• unavailable ipympl, pyqtgraph, mne-qt-browser, ipywidgets, trame_client, trame_server, trame_pyvista, trame_vtk, trame_vuetify

Ecosystem (optional)

• mne-bids 0.18.0
• mne-connectivity 0.8
• neo 0.14.4
• pymef 1.4.8

• unavailable mne-nirs, mne-features, mne-icalabel, mne-bids-pipeline, eeglabio, edfio, curryreader, mffpy, pybv, antio, defusedxm

[Gnefil]

Hey @wmvanvliet, would you be interested in looking into this?

[wmvanvliet]

I already did, see #13955

[wmvanvliet]

This is not really a bug, but an inconsistency in default blocking behavior. evoked.plot_topomap does not block by default, allowing stc.plot() to continue. However, mne.viz.plot_evoked_topo does block by default, preventing any further event handling to take place. In your example, things can be fixed by starting the script with plt.ion(), which disables blocking, and ending it with plt.show(block=True) which re-enables blocking (after all three figures have been created).

[Gnefil]

I see; now I understand the underlying logic. So it wasn't meant for scripts to pop up persistent figures, unlike IPython.
Then, the expected behaviour is to rapidly open and close for all the figures under script mode, unless plot.ion() + plt.show(block=True). I could close the issue once the discussion and strategy correctly implemented, if that makes sense.