Open Ephys: find root folder to correctly look for timestamps

[alejoe91]

Fixes #4496

[alejoe91]

@bparks13 this should fix the issue. It also covers the case where one passes an individual experiment/recording

[chrishalcrow]

Looks good - works for my locally. Could add a test here:

spikeinterface/src/spikeinterface/extractors/tests/test_neoextractors.py

Line 158 in 89381fe

something like:

def check_timestamps_loaded_from_subfolder(self):
"""
Checks if the timestamps are loaded even if user inputs a subfolder as the data path.
"""

folder_path = local_folder / r"openephysbinary/v0.6.x_neuropixels_multiexp_multistream/Record Node 101/experiment1/recording1"
recording = self.ExtractorClass(folder_path, load_sync_timestamps=True)
assert recording.get_time_info().get('time_vector') is not None

(I wrote this in the github box... treat as pseudocode!)

[bparks13]

@alejoe91 Perfect, thank you. +1 for adding checks for the various levels of folder paths that can be given, I didn't even think about that.

@chrishalcrow If the load_sync_timestamps=True parameter is not set, then internally the time vector is not created? That definitely makes it easier to add a unit test since we don't need to add a new dataset. I assumed that it would create a time vector using the start time and the assumed sampling frequency, but it sounds like that is not the case. Or, at the very least, there is a difference in the data structure that we can leverage here.

[chrishalcrow]

@chrishalcrow If the load_sync_timestamps=True parameter is not set, then internally the time vector is not created? That definitely makes it easier to add a unit test since we don't need to add a new dataset. I assumed that it would create a time vector using the start time and the assumed sampling frequency, but it sounds like that is not the case. Or, at the very least, there is a difference in the data structure that we can leverage here.

Yeah, the time vector is pretty huge since it's the length of the total samples of the recording. So if we're using the "start time + sampling frequency" method, we never make it!