Fix: Handle inline backends in plt_show

[mh105]

Reference issue (if any)

None

What does this implement/fix?

mne.viz.utils.plt_show() currently calls Figure.show() when it receives a figure. With a Matplotlib inline backend, that call does not reliably trigger display in notebook-style frontends. This includes the VS Code Interactive Window that some people use. matplotlib.pyplot.show() does.

This change detects inline backend identifiers case-insensitively and routes them through plt.show(**kwargs). Other GUI backends continue to use (fig or plt).show(**kwargs), while the existing no-show behavior for Agg and show=False remains unchanged.

Additional information

I manually identified the issue and implemented the fix, but Codex added the test case to mne/viz/tests/test_utils.py

[larsoner]

I'm okay with this at the MNE-Python end, but it does seem like a workaround... is there an upstream issue? Would be good to let them know this is an issue if they don't

In the meantime CIs have failed due to style (line too long), can you fix it?

[mh105]

@larsoner I actually think this is an appropriate MNE-Python fix:

matplotlib "inline" backend is meant to be non-interactive similar to "agg", but that just means we can't manage the figure handle after its creation. plt.show() is a pyplot-level operation that displays all figures however the backend created them would display them.

I think the main issue here is that we always call fig.show(**kwargs) whenever the backend is not "agg" and the fig handle is not None in

(fig or plt).show(**kwargs)

But backend != "agg" doesn't cover these other backends that don't expose Figure-level manager but are nevertheless not headless rendering backends like Agg.

For this reason, I didn't want to add a condition to if show and backend != "agg":, because despite being non-interactive, plt.show(**kwargs) can still display non-interactive figures created with the "inline" backend. So far this is the only matplotlib-native displayable yet non-interactive backend that I can think of, hence the str search. If we find more in the future, perhaps we can handle this properly and modify the outer if loop all together (if show and backend != "agg":).

[mh105]

I edited the comment line as well. The still failing CI seems to be out of my control...

[wmvanvliet]

Could you register for an Circle-CI account so that the documentation CI can run?

[larsoner]

Thanks @mh105 !