Fix evoked topomap colorbars, closes #13050

[ruuskas]

This should fix #13050 as discussed there.

It now works like this:

1. If you don't pass vlim or contours, MNE sets vlim using max(abs(...)) then chooses contours within those bounds
2. If you pass vlim but not contours, MNE chooses contours to be within the vlim
3. If you pass contours but not vlim, MNE chooses vlim to encompass the contours and max(abs(...))
4. If you pass both contours and vlim, MNE uses them as-is (and if this makes the colorbar have gaps, that's okay/correct)

1. If you don't pass vlim or contours, MNE sets vlim using max(abs(...)) then chooses contours within those bounds

times = np.arange(0.05, 0.151, 0.02)
fig = evoked.plot_topomap(times, ch_type="mag")

2. If you pass vlim but not contours, MNE chooses contours to be within the vlim

fig = evoked.plot_topomap(times, ch_type="mag", vlim=(-400, 400))

3. If you pass contours but not vlim, MNE chooses vlim to encompass the contours and max(abs(...))

fig = evoked.plot_topomap(times, ch_type="mag", contours=[-600, -300, 0, 300, 600])

4. If you pass both contours and vlim, MNE uses them as-is (and if this makes the colorbar have gaps, that's okay/correct)

fig = evoked.plot_topomap(times, ch_type="mag", 
                          contours=[-600, -300, 0, 300, 600], 
                          vlim=(-400, 400))

This also fixes the issue with plot_evoked_joint:

fig = evoked.plot_joint(times=[0.05, 0.15, 0.2], picks='mag')

WDYT?

[larsoner]

Wow a pretty small diff!

@ruuskas can you make some tiny change to examples/visualization/evoked_topomap.py (can just be a reflow or newline add/remove)? That way CircleCI will build the example (it looks for changed ones by default and only runs those) and we can see the results. Feel free to also touch any other examples that you might know offhand look bad currently and might magically be fixed by this PR. That way we don't have to run them manually on our machines (or worse, wait until the PR lands in main and look at the deployed doc) to see what changes.

Also needs a doc/changes/devel/13064.bugfix.rst mentioning the change.

Can you add the "it now works like this:" 4-point conditional to the Notes section of plot_evoked_topomap? I think it's a good place to make clear the default behavior

[ruuskas]

It seems that something is broken with the CIs @larsoner @drammock.

[drammock]

https://app.circleci.com/pipelines/github/mne-tools/mne-python/26410/workflows/72bf256a-8e09-444e-b46e-532f1ec83809/jobs/71029/parallel-runs/0/steps/0-142?invite=true#step-142-83999_21

evoked.plot_topomap docstring has a bullet list that ends without a blank line, according to the warning message

[ruuskas]

https://app.circleci.com/pipelines/github/mne-tools/mne-python/26410/workflows/72bf256a-8e09-444e-b46e-532f1ec83809/jobs/71029/parallel-runs/0/steps/0-142?invite=true#step-142-83999_21

evoked.plot_topomap docstring has a bullet list that ends without a blank line, according to the warning message

I see that now. Earlier today the CircleCI run wouldn't start at all.

[larsoner]

@ruuskas looks good but I think this point above is still pending:

@ruuskas can you make some tiny change to examples/visualization/evoked_topomap.py (can just be a reflow or newline add/remove)? That way CircleCI will build the example (it looks for changed ones by default and only runs those) and we can see the results. Feel free to also touch any other examples that you might know offhand look bad currently and might magically be fixed by this PR. That way we don't have to run them manually on our machines (or worse, wait until the PR lands in main and look at the deployed doc) to see what changes.

[ruuskas]

@ruuskas looks good but I think this point above is still pending:

@ruuskas can you make some tiny change to examples/visualization/evoked_topomap.py (can just be a reflow or newline add/remove)? That way CircleCI will build the example (it looks for changed ones by default and only runs those) and we can see the results. Feel free to also touch any other examples that you might know offhand look bad currently and might magically be fixed by this PR. That way we don't have to run them manually on our machines (or worse, wait until the PR lands in main and look at the deployed doc) to see what changes.

Hmm I did add just a newline in the end but it didn't render the plots. I'll try again.

[larsoner]

Hmm I did add just a newline in the end but it didn't render the plots. I'll try again.

Hah -- that's a style error, so pre-commit.ci fixed that "error" for you!

[ruuskas]

Now it seems to have worked and the colorbars look ok.

[larsoner]

Thanks @ruuskas