Fix SetRawAnnotations when no stim channel

[PierreGtch]

As described by @simonkojima in #830, when a dataset does not have a stim channel, the paradigm ignores the dataset's interval attribute when creating windows

This PR should fix it

@simonkojima could you review this? On my machine, it passes the test on Dreyer2023 you shared

Fixes #830

[simonkojima]

Thank you, @PierreGtch.
It gives different data depending on dataset.interval.

However, I’m now facing a new issue related to missing events.
Each subject should have 240 events, but the current output returns fewer events (e.g., only 11 for subject 1).

It seems to be caused by the commit in this PR, as the issue does not occur on the current develop branch.

from pathlib import Path
import mne
from moabb.datasets import Dreyer2023
from moabb.paradigms import LeftRightImagery

paradigm = LeftRightImagery()

dataset = Dreyer2023()
X, labels, meta = paradigm.get_data(dataset=dataset, subjects=[1])

print(X.shape[0]) # returns 11 (should be 240)

# extract events from raw processed by moabb
raw = dataset.get_data(subjects=[1])[1]["0"]["0R1acquisition"]
events_moabb, _ = mne.events_from_annotations(raw)

# extract events from unprocessed raw
fname = Path.home() / "mne_data" / "MNE-Dreyer2023-data" / "sub-01" / "eeg" / "sub-01_task-R1acquisition_eeg.edf"
raw = mne.io.read_raw_edf(fname)
events, event_id = mne.events_from_annotations(raw)
new_event_id = {"left_hand":event_id["769"], "right_hand":event_id["770"]}
epochs = mne.Epochs(raw, event_id = new_event_id, tmin=0, tmax=5, baseline=None)
events_unprocessed = epochs.events

print(events_moabb.shape) # 2, 3
print(events_unprocessed.shape) # 40, 3

[PierreGtch]

Ok i think I got it!
The final solution is quite convoluted: basically when no stim channel is present, we do annotations -> events -> annotations -> events
Your code works on my machine. I get :

240
(40, 3)
(40, 3)

Feel free to do additional tests, I will merge when tell me
Thanks for your help :)

[simonkojima]

Thanks! I tested it on my machine, and everything works as expected.

[PierreGtch]

Great! thanks @simonkojima :)