mne.viz.plot_compare_evokeds should not draw colorbar legend when legend=False

[mmagnuski]

When a colormap is passed to mne.viz.plot_compare_evokeds (for example cbar='viridis') it places a colorbar next to the axis (instead of a normal in-axis legend). While the presence of the standard legend can be controlled with legend=False, the colorbar legend is always drawn.

Example:

import numpy as np
import mne

info = mne.create_info(ch_names=list('ABCDE'), sfreq=50., ch_types='eeg')

conditions = ['cond1', 'cond2']
evokeds = dict()
for condition in conditions:
    data = np.random.random((5, 100)) * 1e-6
    evokeds[condition] = mne.EvokedArray(data, info, tmin=-0.25)

legend=False controls whether standard legend is drawn:

mne.viz.plot_compare_evokeds(evokeds)

mne.viz.plot_compare_evokeds(evokeds, legend=False)

However, when cbar argument is used, the colormap legend is always drawn:

mne.viz.plot_compare_evokeds(evokeds, cmap='viridis', legend=False)

I think legend=False should not plot the colorbar.

[hoechenberger]

I didn't know that feature existed! Why do we even have that? They're both essentially doing the same, no??

[mmagnuski]

The colorbar feature might be useful for variables with continuous meaning (contrast for example). I think this scenario is assumed when you specify a colormap with cmap.

[mmagnuski]

take a look at this example for instance:
https://mne.tools/stable/auto_examples/stats/sensor_regression.html

[hoechenberger]

Makes sense!

[drammock]

hmm. interestingly mne.viz.plot_compare_evokeds(evokeds, cmap='viridis', legend=True) does not show the in-axes legend. It's been a while since I've touched this code... does it make sense that if you specify cmap then you will never get an in-axes legend? If so, then it could be safe to use legend to control in-axes legend when cmap=None and control colorbar when cmap is not None.