Reading continuous data with cbpy

[jsv13]

I manage to retrieve events from nPlayServer with cbpy, but I cannot get continuous data. Streaming continuous data is on, e.g. I can get them via the cbSDK.
Can anyone provide an example showing how to get continuous data in cbpy?
thanks!

[cboulay]

I wrote a convenience class for myself that wraps cbpy that I use in a few different applications.
You can see how I get continuous data here.

[zeeMonkeez]

@cboulay how do you match sample times between events and continuous data in this use case? cbMex returns a common time base, but it's not clear to me how cbpy might expose that...

[cboulay]

@zeeMonkeez Hi Jonas. Sorry I can't look into it right now. (going to Asilomar? I am nowhere near ready)

Anyway, there's a recently contributed get_data function in the 7.0 branch here. This fetches events and continuous data together in one function after a reset. The timestamps of the events should be offset from the 0th sample of the continuous data.

I haven't had a need for this because my I'm either looking at these things independently, or I'm looking at them offline loaded from .nev/.nsx files, but I think that's going to change.

[zeeMonkeez]

@cboulay Hey Chad, thanks I'll have a look. No, Asilomar didn't fit in my schedule. I'll be in Minneapolis in June, though, will you be there? Seems like you're up to doing some cool things, it'd be cool to chat...

[cboulay]

@jsv13 Very late reply, but in case anyone else encounters this, I've noticed that channels coming from the Analog Inputs aren't replayed by nPlayServer.

[stinos]

I've noticed that channels coming from the Analog Inputs aren't replayed by nPlayServer.

According to Blackrock support this is a known bug in the current version 7, a fix is upcoming, and it should also not happen in nPlayServer versions 6.5.4 and below.

[cboulay]

Closing. Use #97

[agamemnonc]

I am also trying to read continuous data only using cbpy and have encountered the following issue.

I have a main program which queries the device for new data continuously (i.e., no sleep commands in-between). I have set the continuous_length in buffer_parameter to the desired samples I want to read per every read operation.

from cerebus import cbpy
samples_per_read = 100

buffer_parameter = {
    'double': True,
    'continuous_length': samples_per_read 
}
result, _ = cbpy.trial_config(
    reset=True,
    buffer_parameter=buffer_parameter)

def read():
    result, trial = cbpy.trial_continuous(reset=True)
    data = process(trial)
    return data

Now, ideally I would want to continuously read data in a loop, such that execution is blocked until the new data become available, something like:

while True:
   data = read()

I would expect program execution to be blocked until data become available since continuous_length is set to the desired number of samples per read operation, but what I rather observe is that a smaller number of samples is read and after a little bit, when no samples are available , an exception is raised. Is this expected behaviour?

Alternatively, I could set the number of samples per read to 1 and internally check if the desired samples have been read, but querying the device sample by sample sounds a bit like an overkill.

Any ideas? Thanks.

[cboulay]

It would be a non-trivial amount of work to add a timeout parameter to SdkGetTrialData. It's tricky to do correctly and I don't have time for that right now.

The simplest solution is to buffer the data on the client side. Or, better yet, have your analysis pipeline be flexible to the number of samples it receives (ultimately that usually means buffering at some stage).

If you want to make an argument for a blocking call to fetch data or you want to propose a way to do that then please open a new issue.