Recordings

Opening a file reads only its metadata. Samples are read when you ask for them.

import dspk

rec = dspk.io.open("session.zarr")
rec.channel_count, rec.samples, rec.sample_rate, rec.duration
rec.channel_names, rec.units

first_second = rec.read(start=0, stop=30_000)              # all channels
some = rec.read(channels=[3, 0], start=0, stop=30_000)     # in the order given

read returns (channels, samples) float32 in each channel’s unit.

Files with several sources

Containers such as NWB hold several recordings. List them first:

for s in dspk.io.sources("session.nwb.zarr"):
    print(s.id, s.channel_count, s.sample_rate)

rec = dspk.io.open("session.nwb.zarr", source="acquisition/EMG")

Writing

dspk.io.write_zarr(data, "out.zarr", 30_000.0, unit="µV")              # float32
dspk.io.write_raw(counts, "out.bin", 30_000.0, gain=0.195, unit="µV")  # integers kept as they are

For data that doesn’t fit in memory, ZarrWriter takes blocks one at a time.

dspk.io.formats() lists what open recognizes. Neural formats (SpikeGLX, NWB) come with the Synapse plugin.

Full details: dspk.io.