def test_multidimensional_features():
"""Features should be split into sub-features when they
are multidimensional.
This should be the case even when the feature is `None` or `[]`
The following neuron has no axon but the axon feature segment_midpoints for
the axon should still be made of 3 values (X, Y and Z)
Cf: https://github.com/BlueBrain/NeuroM/issues/859
"""
neuron = nm.load_neuron(Path(SWC_PATH, 'no-axon.swc'))
config = {
'neurite': {
'segment_midpoints': ['max']
},
'neurite_type': ['AXON']
}
actual = ms.extract_dataframe(neuron, config)
assert_array_equal(
actual['axon'][[
'max_segment_midpoint_0', 'max_segment_midpoint_1',
'max_segment_midpoint_2'
]].values, [[None, None, None]])
config = {'neurite': {'partition_pairs': ['max']}}
actual = ms.extract_dataframe(neuron, config)
assert_array_equal(
actual['axon'][['max_partition_pair_0',
'max_partition_pair_1']].values, [[None, None]])
def test_extract_dataframe_multiproc():
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns, REF_CONFIG, n_workers=2)
expected = pd.read_csv(Path(DATA_PATH, 'extracted-stats.csv'), index_col=0, header=[0, 1])
assert_frame_equal(actual, expected)
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns, REF_CONFIG, n_workers=os.cpu_count() + 1)
assert_equal(len(w), 1, "Warning not emitted")
assert_frame_equal(actual, expected)
def features(self, config: Dict, n_workers=1):
'''Returns a dataframe containing morphometrics and neurondb information
Args:
config: a NeuroM morph_stas config.
See https://neurom.readthedocs.io/en/latest/morph_stats.html for more information
n_workers: the number of workers to use to perform the computations
Returns:
A jointure dataframe between `neurom.stats.extract_dataframe` and `self.df`
Raises:
ValueError: if `self.df` has undefined (ie. None) paths
'''
missing_morphs = self.df[self.df.path.isnull()].name.to_list()
if missing_morphs:
raise ValueError(
f'DataFrame has morphologies with undefined filepaths: {missing_morphs}'
)
df = self.df.copy().reset_index(drop=True)
df.columns = pd.MultiIndex.from_product(
(["properties"], df.columns.values))
stats = extract_dataframe(df['properties', 'path'], config, n_workers)
return df.join(stats.drop(columns='name', level=1), how='inner')
def test_extract_dataframe_multiproc():
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns, REF_CONFIG, n_workers=2)
expected = pd.read_csv(Path(DATA_PATH, 'extracted-stats.csv'), index_col=0)
# Compare sorted DataFrame since Pool.imap_unordered disrupted the order
assert_frame_equal(actual.sort_values(by=['name']).reset_index(drop=True),
expected.sort_values(by=['name']).reset_index(drop=True))
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns, REF_CONFIG, n_workers=os.cpu_count() + 1)
assert_equal(len(w), 1, "Warning not emitted")
assert_frame_equal(actual.sort_values(by=['name']).reset_index(drop=True),
expected.sort_values(by=['name']).reset_index(drop=True))
def test_extract_dataframe_multiproc():
# FIXME: Cannot use Neuron objects in the extract_dataframe ctor right now
# because of "TypeError: can't pickle Neuron objects"
# nrns = nm.load_neurons([Path(SWC_PATH, name)
# for name in ['Neuron.swc', 'simple.swc']])
nrns = [Path(SWC_PATH, name) for name in ['Neuron.swc', 'simple.swc']]
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns, REF_CONFIG, n_workers=2)
expected = pd.read_csv(Path(DATA_PATH, 'extracted-stats.csv'),
index_col=0,
header=[0, 1])
assert_frame_equal(actual, expected)
with warnings.catch_warnings(record=True) as w:
actual = ms.extract_dataframe(nrns,
REF_CONFIG,
n_workers=os.cpu_count() + 1)
assert len(w) == 1, "Warning not emitted"
assert_frame_equal(actual, expected)
def test_extract_dataframe():
# Vanilla test
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
actual = ms.extract_dataframe(nrns, REF_CONFIG)
expected = pd.read_csv(Path(DATA_PATH, 'extracted-stats.csv'), header=[0, 1], index_col=0)
assert_frame_equal(actual, expected)
# Test with a single neuron in the population
nrns = nm.load_neurons(Path(SWC_PATH, 'Neuron.swc'))
actual = ms.extract_dataframe(nrns, REF_CONFIG)
assert_frame_equal(actual, expected.iloc[[0]], check_dtype=False)
# Test with a config without the 'neuron' key
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
config = {'neurite': {'section_lengths': ['total']},
'neurite_type': ['AXON', 'APICAL_DENDRITE', 'BASAL_DENDRITE', 'ALL']}
actual = ms.extract_dataframe(nrns, config)
idx = pd.IndexSlice
expected = expected.loc[:, idx[:, ['name', 'total_section_length']]]
assert_frame_equal(actual, expected)
# Test with a FstNeuron argument
nrn = nm.load_neuron(Path(SWC_PATH, 'Neuron.swc'))
actual = ms.extract_dataframe(nrn, config)
assert_frame_equal(actual, expected.iloc[[0]], check_dtype=False)
# Test with a List[FstNeuron] argument
nrns = [nm.load_neuron(Path(SWC_PATH, name))
for name in ['Neuron.swc', 'simple.swc']]
actual = ms.extract_dataframe(nrns, config)
assert_frame_equal(actual, expected)
# Test with a List[Path] argument
nrns = [Path(SWC_PATH, name) for name in ['Neuron.swc', 'simple.swc']]
actual = ms.extract_dataframe(nrns, config)
assert_frame_equal(actual, expected)
# Test without any neurite_type keys, it should pick the defaults
config = {'neurite': {'total_length_per_neurite': ['total']}}
actual = ms.extract_dataframe(nrns, config)
expected_columns = pd.MultiIndex.from_tuples(
[('property', 'name'),
('axon', 'total_total_length_per_neurite'),
('basal_dendrite', 'total_total_length_per_neurite'),
('apical_dendrite', 'total_total_length_per_neurite'),
('all', 'total_total_length_per_neurite')])
expected = pd.DataFrame(
columns=expected_columns,
data=[['Neuron', 207.87975221, 418.43241644, 214.37304578, 840.68521442],
['simple', 15., 16., 0., 31., ]])
assert_frame_equal(actual, expected)
def test_extract_dataframe():
# Vanilla test
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
actual = ms.extract_dataframe(nrns, REF_CONFIG)
expected = pd.read_csv(Path(DATA_PATH, 'extracted-stats.csv'), index_col=0)
assert_frame_equal(actual, expected)
# Test with a single neuron in the population
nrns = nm.load_neurons(Path(SWC_PATH, 'Neuron.swc'))
actual = ms.extract_dataframe(nrns, REF_CONFIG)
assert_frame_equal(actual, expected[expected.name == 'Neuron'], check_dtype=False)
# Test with a config without the 'neuron' key
nrns = nm.load_neurons([Path(SWC_PATH, name)
for name in ['Neuron.swc', 'simple.swc']])
config = {'neurite': {'section_lengths': ['total']},
'neurite_type': ['AXON', 'APICAL_DENDRITE', 'BASAL_DENDRITE', 'ALL']}
actual = ms.extract_dataframe(nrns, config)
expected = expected[['name', 'neurite_type', 'total_section_length']]
assert_frame_equal(actual, expected)
# Test with a FstNeuron argument
nrn = nm.load_neuron(Path(SWC_PATH, 'Neuron.swc'))
actual = ms.extract_dataframe(nrn, config)
assert_frame_equal(actual, expected[expected.name == 'Neuron'], check_dtype=False)
# Test with a List[FstNeuron] argument
nrns = [nm.load_neuron(Path(SWC_PATH, name))
for name in ['Neuron.swc', 'simple.swc']]
actual = ms.extract_dataframe(nrns, config)
assert_frame_equal(actual, expected)
# Test with a List[Path] argument
nrns = [Path(SWC_PATH, name) for name in ['Neuron.swc', 'simple.swc']]
actual = ms.extract_dataframe(nrns, config)
assert_frame_equal(actual, expected)
# Test without any neurite_type keys, it should pick the defaults
config = {'neurite': {'total_length_per_neurite': ['total']}}
actual = ms.extract_dataframe(nrns, config)
expected = pd.DataFrame(
columns=['name', 'neurite_type', 'total_total_length_per_neurite'],
data=[['Neuron', 'axon', 207.879752],
['Neuron', 'basal_dendrite', 418.432416],
['Neuron', 'apical_dendrite', 214.373046],
['Neuron', 'all', 840.685214],
['simple', 'axon', 15.000000],
['simple', 'basal_dendrite', 16.000000],
['simple', 'apical_dendrite', 0.000000],
['simple', 'all', 31.000000]])
assert_frame_equal(actual, expected)
