OpenEphysBinaryRawIO: Fixing ttl multichan

[vigji]

Addressing #1437.

Currently, all the logic in the events parser for OpenEphys data assumes that there is a single stream of events (simply positive or negative states), and it breaks down when acquiring multiple digital signals at once (states having different absolute values for different channels).

This is a simple fix that looks for rising and falling edges separately for each of the digital channels.

Add simple testing relying on existing test data, which fails before the fix and passes afterwards.

[alejoe91]

Thanks @vigji! Looking great!

[vigji]

Great, let me know if there's anything needed before merging!

[zm711]

This looks good to me, although I would ask for one additional comment in the code to explain what version this works for (all or v0.6+ whatever) along with what the structure of the data is.

[zm711]

This works for me, but Alessio knows this way better so I'll let him decide when to merge/do final review.

[alejoe91]

See comment

stale

[samuelgarcia]

Hi,
I was totally off for a while.
I read this very very quickly.
I suspect that this np.where() can slow down a lot the parse_header for long recording.
Is it a general or a corner case ?
We need to keep in mind that the parse header is done multiple time when multiprocessing on spikeinterface side so we need to be very carfull for this.
Do we have benchmark for this ?

[vigji]

Not data but rule of thumbs benchmark:
assuming 10 channels sampling TTL pulses at 100 Hz for 5 hours I generate a random array of integers from 0 to 10, np.where() then takes approx. 15 ms on my laptop (MacBook Pro M1 2022). This matches my experience that I never see long times even for fast digital signals recorded for a couple of hours.

Does this make sense?

[samuelgarcia]

then it is ok. thank you for the fast anwser

[zm711]

@vigji do we have you in our author list? I'm prepping our documents for new release. Would you like me to add you to the list (and what is the institution you were at/or are at for your Neo contribution?)

[vigji]

I don't think so as it was my first contribution!
Yes I'd like to, my institute is [optional: Center for Neuroscience and Cognitive systems], Istituto Italiano di Tecnologia (IIT), Italy

[zm711]

Perfect. I will add you before our next release :) Thanks for the contribution!