Fix Brain slider ranges

[wmvanvliet]

This updates the logic that determines the ranges for the three sliders (fmin, fmid, fmax) on the left of a Brain figure. There was an overflow when the data is all zeros (np.log10(fmax)) which this PR fixes by using np.log10(max(fmax, 1e-30)) instead. While I was there, I noticed that the minimum and maximum values for the sliders were set based on the fmin and fmax of the current timepoint instead of the entire data. Finally, on second thought, the minimum value for the sliders should always be 0, that makes a lot more sense to me than capping it to fmin.

To check out the new behavior:

import mne
path = mne.datasets.sample.data_path()
stc = mne.read_source_estimate(path / "MEG/sample/fsaverage_audvis-meg")
brain = stc.plot("fsaverage", hemi="both", subjects_dir=path / "subjects")

fixes #12606

[drammock]

behavior does seem better, thanks! Please add a changelog entry.

[drammock]

once #12615 lands we can merge main and that should get CIs green here

[hoechenberger]

i love it 😍

[larsoner]

@wmvanvliet looks like this broke an example:

https://app.circleci.com/pipelines/github/mne-tools/mne-python/24096/workflows/69ebaf39-fe26-4774-83e0-238a1fd85918/jobs/66162

Unexpected failing examples:

    ../examples/inverse/source_space_snr.py failed leaving traceback:

    Traceback (most recent call last):
      File "/home/circleci/project/examples/inverse/source_space_snr.py", line 74, in <module>
        brain = snr_stc.plot(**kwargs)
      ...
      File "/home/circleci/project/mne/viz/_brain/_brain.py", line 4164, in _get_range
        fscale_power = int(np.log10(max(fmax, np.finfo("float32").min)))
    RuntimeWarning: invalid value encountered in log10

And locally just running the example I get the warning then an error:

/home/larsoner/python/mne-python/mne/viz/_brain/_brain.py:4164: RuntimeWarning: invalid value encountered in log10
  fscale_power = int(np.log10(max(fmax, np.finfo("float32").min)))
Traceback (most recent call last):
  File "/home/larsoner/python/mne-python/examples/inverse/source_space_snr.py", line 74, in <module>
    brain = snr_stc.plot(**kwargs)
            ^^^^^^^^^^^^^^^^^^^^^^
  ...
  File "/home/larsoner/python/mne-python/mne/viz/_brain/_brain.py", line 4164, in _get_range
    fscale_power = int(np.log10(max(fmax, np.finfo("float32").min)))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[wmvanvliet]

Oh no! That looks like a platform issue. Was np.finfo("float32").min a poor choice?

[wmvanvliet]

Yes, I really think it was a poor choice. I meant to do this: #12619