def test_stimulus_file_from_json(stimulus_file_fixture):
stim_pkl_path, stim_pkl_data = stimulus_file_fixture
# Basic test case
input_json_dict = {"behavior_stimulus_file": str(stim_pkl_path)}
stimulus_file = StimulusFile.from_json(input_json_dict)
assert stimulus_file.data == stim_pkl_data
# Now test caching by deleting the stimulus_file
stim_pkl_path.unlink()
stimulus_file_cached = StimulusFile.from_json(input_json_dict)
assert stimulus_file_cached.data == stim_pkl_data
def test_get_licks_excess(self, tmpdir):
"""
Test that Licks.from_stimulus_file
in the case where
there is an extra frame at the end of the trial log and the mouse
licked on that frame
https://github.com/AllenInstitute/visual_behavior_analysis/blob
/master/visual_behavior/translator/foraging2/extract.py#L640-L647
"""
stimulus_filepath = self._create_test_stimulus_file(
lick_events=[12, 15, 90, 136, 200], # len(timestamps) == 200,
tmpdir=tmpdir)
stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(stimulus_filepath)})
timestamps = StimulusTimestamps(timestamps=np.arange(0, 2.0, 0.01))
licks = Licks.from_stimulus_file(stimulus_file=stimulus_file,
stimulus_timestamps=timestamps)
expected_dict = {
'timestamps': [0.12, 0.15, 0.90, 1.36],
'frame': [12, 15, 90, 136]
}
expected_df = pd.DataFrame(expected_dict)
assert expected_df.columns.equals(licks.value.columns)
np.testing.assert_array_almost_equal(
expected_df.timestamps.to_numpy(),
licks.value['timestamps'].to_numpy(),
decimal=10)
np.testing.assert_array_almost_equal(expected_df.frame.to_numpy(),
licks.value['frame'].to_numpy(),
decimal=10)
def test_from_stimulus_file2(self, tmpdir):
"""
Test that Licks.from_stimulus_file returns a dataframe
of licks whose timestamps are based on their frame number
with respect to the stimulus_timestamps
"""
stimulus_filepath = self._create_test_stimulus_file(
lick_events=[12, 15, 90, 136], tmpdir=tmpdir)
stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(stimulus_filepath)})
timestamps = StimulusTimestamps(timestamps=np.arange(0, 2.0, 0.01))
licks = Licks.from_stimulus_file(stimulus_file=stimulus_file,
stimulus_timestamps=timestamps)
expected_dict = {
'timestamps': [0.12, 0.15, 0.90, 1.36],
'frame': [12, 15, 90, 136]
}
expected_df = pd.DataFrame(expected_dict)
assert expected_df.columns.equals(licks.value.columns)
np.testing.assert_array_almost_equal(
expected_df.timestamps.to_numpy(),
licks.value['timestamps'].to_numpy(),
decimal=10)
np.testing.assert_array_almost_equal(expected_df.frame.to_numpy(),
licks.value['frame'].to_numpy(),
decimal=10)
def from_json(cls, dict_repr: dict) -> "StimulusTimestamps":
if 'sync_file' in dict_repr:
sync_file = SyncFile.from_json(dict_repr=dict_repr)
return cls.from_sync_file(sync_file=sync_file)
else:
stim_file = StimulusFile.from_json(dict_repr=dict_repr)
return cls.from_stimulus_file(stimulus_file=stim_file)
def from_json(cls,
session_data: dict,
monitor_delay: Optional[float] = None) \
-> "BehaviorSession":
"""
Parameters
----------
session_data
Dict of input data necessary to construct a session
monitor_delay
Monitor delay. If not provided, will use an estimate.
To provide this value, see for example
allensdk.brain_observatory.behavior.data_objects.stimuli.util.
calculate_monitor_delay
Returns
-------
`BehaviorSession` instance
"""
behavior_session_id = BehaviorSessionId.from_json(
dict_repr=session_data)
stimulus_file = StimulusFile.from_json(dict_repr=session_data)
stimulus_timestamps = StimulusTimestamps.from_json(
dict_repr=session_data)
running_acquisition = RunningAcquisition.from_json(
dict_repr=session_data)
raw_running_speed = RunningSpeed.from_json(dict_repr=session_data,
filtered=False)
running_speed = RunningSpeed.from_json(dict_repr=session_data)
metadata = BehaviorMetadata.from_json(dict_repr=session_data)
if monitor_delay is None:
monitor_delay = cls._get_monitor_delay()
licks, rewards, stimuli, task_parameters, trials = \
cls._read_data_from_stimulus_file(
stimulus_file=stimulus_file,
stimulus_timestamps=stimulus_timestamps,
trial_monitor_delay=monitor_delay
)
date_of_acquisition = DateOfAcquisition.from_json(
dict_repr=session_data)\
.validate(
stimulus_file=stimulus_file,
behavior_session_id=behavior_session_id.value)
return BehaviorSession(behavior_session_id=behavior_session_id,
stimulus_timestamps=stimulus_timestamps,
running_acquisition=running_acquisition,
raw_running_speed=raw_running_speed,
running_speed=running_speed,
metadata=metadata,
licks=licks,
rewards=rewards,
stimuli=stimuli,
task_parameters=task_parameters,
trials=trials,
date_of_acquisition=date_of_acquisition)
def test_get_licks_failure(self, tmpdir):
stimulus_filepath = self._create_test_stimulus_file(
lick_events=[12, 15, 90, 136, 201], # len(timestamps) == 200,
tmpdir=tmpdir)
stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(stimulus_filepath)})
timestamps = StimulusTimestamps(timestamps=np.arange(0, 2.0, 0.01))
with pytest.raises(IndexError):
Licks.from_stimulus_file(stimulus_file=stimulus_file,
stimulus_timestamps=timestamps)
def test_stimulus_timestamps_from_json2():
dir = Path(__file__).parent.parent.resolve()
test_data_dir = dir / 'test_data'
sf_path = test_data_dir / 'stimulus_file.pkl'
sf = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(sf_path)})
stimulus_timestamps = StimulusTimestamps.from_stimulus_file(
stimulus_file=sf)
expected = np.array([0.016 * i for i in range(11)])
assert np.allclose(expected, stimulus_timestamps.value)
def test_from_stimulus_file2(self, tmpdir):
"""
Test that Rewards.from_stimulus_file returns
expected results (main nuance is that timestamps should be
determined by applying the reward frame as an index to
stimulus_timestamps)
"""
def _create_dummy_stimulus_file():
trial_log = [{
'rewards': [(0.001, -1.0, 4)],
'trial_params': {
'auto_reward': True
}
}, {
'rewards': []
}, {
'rewards': [(0.002, -1.0, 10)],
'trial_params': {
'auto_reward': False
}
}]
data = {
'items': {
'behavior': {
'trial_log': trial_log
}
},
}
tmp_path = tmpdir / 'stimulus_file.pkl'
with open(tmp_path, 'wb') as f:
pickle.dump(data, f)
f.seek(0)
return tmp_path
stimulus_filepath = _create_dummy_stimulus_file()
stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(stimulus_filepath)})
timestamps = StimulusTimestamps(timestamps=np.arange(0, 2.0, 0.01))
rewards = Rewards.from_stimulus_file(stimulus_file=stimulus_file,
stimulus_timestamps=timestamps)
expected_dict = {
'volume': [0.001, 0.002],
'timestamps': [0.04, 0.1],
'autorewarded': [True, False]
}
expected_df = pd.DataFrame(expected_dict)
expected_df = expected_df
assert expected_df.equals(rewards.value)
def from_json(cls,
dict_repr: dict,
filtered: bool = True,
zscore_threshold: float = 10.0) -> "RunningSpeed":
stimulus_file = StimulusFile.from_json(dict_repr)
stimulus_timestamps = StimulusTimestamps.from_json(dict_repr)
running_speed = cls._get_running_speed_df(stimulus_file,
stimulus_timestamps,
filtered, zscore_threshold)
return cls(running_speed=running_speed,
stimulus_file=stimulus_file,
stimulus_timestamps=stimulus_timestamps,
filtered=filtered)
def from_json(
cls,
dict_repr: dict,
) -> "RunningAcquisition":
stimulus_file = StimulusFile.from_json(dict_repr)
stimulus_timestamps = StimulusTimestamps.from_json(dict_repr)
running_acq_df = get_running_df(
data=stimulus_file.data,
time=stimulus_timestamps.value,
)
running_acq_df.drop("speed", axis=1, inplace=True)
return cls(
running_acquisition=running_acq_df,
stimulus_file=stimulus_file,
stimulus_timestamps=stimulus_timestamps,
)
def test_empty_licks(self, tmpdir):
"""
Test that Licks.from_stimulus_file in the case where
there are no licks
"""
stimulus_filepath = self._create_test_stimulus_file(lick_events=[],
tmpdir=tmpdir)
stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(stimulus_filepath)})
timestamps = StimulusTimestamps(timestamps=np.arange(0, 2.0, 0.01))
licks = Licks.from_stimulus_file(stimulus_file=stimulus_file,
stimulus_timestamps=timestamps)
expected_dict = {'timestamps': [], 'frame': []}
expected_df = pd.DataFrame(expected_dict)
assert expected_df.columns.equals(licks.value.columns)
np.testing.assert_array_equal(expected_df.timestamps.to_numpy(),
licks.value['timestamps'].to_numpy())
np.testing.assert_array_equal(expected_df.frame.to_numpy(),
licks.value['frame'].to_numpy())
def test_stimulus_timestamps_from_json3():
"""
Test that StimulusTimestamps.from_stimulus_file
just returns the sum of the intervalsms field in the
behavior stimulus pickle file, padded with a zero at the
first timestamp.
"""
dir = Path(__file__).parent.parent.resolve()
test_data_dir = dir / 'test_data'
sf_path = test_data_dir / 'stimulus_file.pkl'
sf = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(sf_path)})
sf._data['items']['behavior']['intervalsms'] = [0.1, 0.2, 0.3, 0.4]
stimulus_timestamps = StimulusTimestamps.from_stimulus_file(
stimulus_file=sf)
expected = np.array([0., 0.0001, 0.0003, 0.0006, 0.001])
np.testing.assert_array_almost_equal(stimulus_timestamps.value,
expected,
decimal=10)
def setup_class(cls):
dir = Path(__file__).parent.parent.parent.resolve()
test_data_dir = dir / 'test_data'
sf_path = test_data_dir / 'stimulus_file.pkl'
cls.stimulus_file = StimulusFile.from_json(
dict_repr={'behavior_stimulus_file': str(sf_path)})