`cut_segment_by_epoch` returns segment with wrong names when using data from NIX files

[kohlerca]

Hello,

I am working with data read from a NIX file using NixIO, which also contain epochs for the individual trials.

However, when I cut the main segment using one epoch, the nix_name annotation in the resulting segment is the same as the Epoch object used to cut the data.

The code below reproduces the behavior with generated data.

import neo
from neo import utils as neo_utils
import numpy as np
import quantities as pq

# Generate a segment with data. Annotations as expected from reading a NIX file

signal = neo.AnalogSignal(np.random.random((90000, 5)), units=pq.uV,
                          sampling_rate=30000*pq.Hz, name="Electrode data",
                          nix_name="neo.analogsignal.ca2b24032c0b4c0497856e24b69268c0",
                          implantation_site="M1")

segment = neo.Segment(name="Segment 0", description="Test segment",
                      nix_name="neo.segment.1dab77a06b734b638f13e9c19d76d6df")

segment.analogsignals = [signal]


# Generate epoch to cut between 1 and 2 s, with annotations

epoch = neo.Epoch([1]*pq.s, [1]*pq.s, name="Epoch 1", description="Test epoch",
                  trial_number=1, trial_protocol=10,
                  nix_name="neo.epoch.da832a0d90c8479abd2247aff4840133")


# Cut the segment using the epoch

trial_segment = neo_utils.cut_segment_by_epoch(segment, epoch)[0]


# Inspect the annotations

print(trial_segment.annotations)

The output is

{'nix_name': 'neo.epoch.da832a0d90c8479abd2247aff4840133', 'trial_number': 1, 'trial_protocol': 10}

One solution would be to store all the available identification attributes of the Epoch object (name, description, nix_name) as a dict in an annotation source_epoch. Or to explicitly ignore the nix_name annotations during the cut.

Thank you,
Cristiano

[JuliaSprenger]

I @kohlerca,
thanks for raising the issue. That is indeed a tricky one as we need to weigh the user-friendliness of automatically copying the annotations from the source epoch to the segment against keeping the code general and not introducing extra rules just for the nix_name annotation.
With cut_segment_by_epoch being just a utility and not a core function I would tend to go for the latter and introduce an special case for the nix_name annotation here. @samuelgarcia @apdavison Do you have stronger opinions here?

[samuelgarcia]

We could make this optional (True by default maybe False ?):

subseg.annotate(**copy.copy(epoch.annotations))

and remove nix_name hardcoded as you propose.

The revelant annoations are mostly this

        # Add array-annotations of Epoch
        for key, val in epoch.array_annotations.items():
            if len(val):
                subseg.annotations[key] = copy.copy(val[ep_id])

[JuliaSprenger]

@samuelgarcia Yes, having it handles by a keyword argument seems the best case.