class TestSingleTypeCompilation(unittest.TestCase):
"""
Check if the scaffold can create a single cell type.
"""
@classmethod
def setUpClass(self):
super(TestSingleTypeCompilation, self).setUpClass()
config = JSONConfig(file=single_neuron_config)
self.scaffold = Scaffold(config)
self.scaffold.compile_network()
def test_placement_statistics(self):
self.assertEqual(self.scaffold.statistics.cells_placed["test_cell"], 4)
self.assertEqual(self.scaffold.get_cell_total(), 4)
def test_network_cache(self):
pass
# TODO: Implement a check that the network cache contains the right amount of placed cells
def test_hdf5_cells(self):
pass
# TODO: Implement a check that the hdf5 file contains the right datasets under 'cells'
def test_from_hdf5(self):
scaffold_copy = from_hdf5(self.scaffold.output_formatter.file)
for key in self.scaffold.statistics.cells_placed:
self.assertEqual(
scaffold_copy.statistics.cells_placed[key],
self.scaffold.statistics.cells_placed[key],
)
self.assertEqual(scaffold_copy.get_cell_total(), 4)
self.assertRaises(OSError, from_hdf5, "doesntexist")
class TestEntities(unittest.TestCase):
@classmethod
def setUpClass(self):
import nest
nest.ResetKernel()
super(TestEntities, self).setUpClass()
config = JSONConfig(file=minimal_config_entities)
self.scaffold = Scaffold(config)
self.scaffold.compile_network()
hdf_config = _from_hdf5("minimal_entities.hdf5")
self.scaffold_fresh = Scaffold(hdf_config,
from_file="minimal_entities.hdf5")
def test_placement_statistics(self):
self.assertEqual(self.scaffold.statistics.cells_placed["entity_type"],
100)
@unittest.skipIf(
importlib.util.find_spec("nest") is None, "NEST is not importable.")
def test_creation_in_nest(self):
f = h5py.File("minimal_entities.hdf5", "r")
ids = list(f["entities"]["entity_type"])
self.assertEqual(ids, list(range(100)))
f.close()
# Try to load the network directly from the hdf5 file
nest_adapter = self.scaffold_fresh.get_simulation("test")
simulator = nest_adapter.prepare()
nest_adapter.simulate(simulator)
nest_ids = nest_adapter.entities["entity_type"].nest_identifiers
self.assertEqual(list(nest_ids), list(range(1, 101)))
class TestSingleNeuronTypeSetup(unittest.TestCase):
def setUp(self):
config = JSONConfig(file=single_neuron_config)
self.scaffold = Scaffold(config)
self.scaffold.compile_network()
self.nest_adapter = self.scaffold.configuration.simulations[
"test_single_neuron"]
self.nest_adapter.reset()
def tearDown(self):
self.nest_adapter.delete_lock()
def test_single_neuron(self):
self.scaffold.run_simulation("test_single_neuron")
test_cell_model = self.nest_adapter.cell_models["test_cell"]
self.assertEqual(test_cell_model.nest_identifiers, list(range(1, 5)))
test_neuron_status = self.nest_adapter.nest.GetStatus(
test_cell_model.nest_identifiers)
self.assertEqual(test_neuron_status[0]["t_ref"], 1.5)
self.assertEqual(test_neuron_status[0]["C_m"], 7.0)
self.assertEqual(test_neuron_status[0]["V_th"], -41.0)
self.assertEqual(test_neuron_status[0]["V_reset"], -70.0)
self.assertEqual(test_neuron_status[0]["E_L"], -62.0)
self.assertEqual(test_neuron_status[0]["I_e"], 0.0)
def test_spoofing(self):
"""
Assert that fake detailed connections can be made
"""
config = JSONConfig(file=_config)
scaffold = Scaffold(config)
scaffold.compile_network()
original_connections = len(
scaffold.cell_connections_by_tag["connection"])
sd = SpoofDetails()
sd.presynaptic = "from_cell"
sd.postsynaptic = "to_cell"
sd.scaffold = scaffold
# Raise error because here's no morphologies registered for the cell types.
with self.assertRaises(
MorphologyDataError,
msg="Missing morphologies during spoofing not caught."):
sd.after_connectivity()
# Add some morphologies
setattr(
config.cell_types["from_cell"].morphology,
"detailed_morphologies",
{"names": ["GranuleCell"]},
)
setattr(
config.cell_types["to_cell"].morphology,
"detailed_morphologies",
{"names": ["GranuleCell"]},
)
# Run the spoofing again
sd.after_connectivity()
cs = scaffold.get_connectivity_set("connection")
scaffold.compile_output()
# Use the intersection property. It throws an error should the detailed
# information be missing
try:
i = cs.intersections
for inter in i:
fl = inter.from_compartment.labels
tl = inter.to_compartment.labels
self.assertIn("axon", fl,
"From compartment type is not an axon")
self.assertIn("dendrites", tl,
"From compartment type is not a dendrite")
self.assertNotEqual(len(i), 0, "Empty intersection data")
self.assertEqual(len(i), original_connections,
"Different amount of spoofed connections")
except MissingMorphologyError:
self.fail("Could not find the intersection data on spoofed set")
# Set both types to relays and try spoofing again
config.cell_types["from_cell"].relay = True
config.cell_types["to_cell"].relay = True
with self.assertRaises(MorphologyError,
msg="Did not catch double relay spoofing!"):
sd.after_connectivity()
def test_representatives(self):
"""
Test that 1 cell per non-relay cell model is chosen.
"""
from patch import p
config = JSONConfig(double_nn_config)
scaffold = Scaffold(config)
scaffold.compile_network()
adapter = scaffold.create_adapter("neuron")
adapter.h = p
adapter.load_balance()
device = adapter.devices["test_representatives"]
device.initialise_targets()
targets = adapter.devices["test_representatives"].get_targets()
self.assertEqual(
1,
len(targets),
"Targetting type `representatives` did not return the correct amount of representatives.",
)
class TestMultiInstance(unittest.TestCase):
@classmethod
def setUpClass(self):
super(TestMultiInstance, self).setUpClass()
import nest
self.nest = nest
config = JSONConfig(file=single_neuron_config)
self.scaffold = Scaffold(config)
self.scaffold.compile_network()
self.hdf5 = self.scaffold.output_formatter.file
self.nest_adapter_0 = self.scaffold.get_simulation(
"test_single_neuron")
# When another test errors, the lock might remain, and all locking tests fail
self.nest_adapter_0.delete_lock()
self.nest_adapter_1 = self.scaffold.create_adapter(
"test_single_neuron")
self.nest_adapter_2 = self.scaffold.create_adapter(
"test_single_neuron")
self.nest_adapter_multi_1 = self.scaffold.create_adapter(
"test_single_neuron")
self.nest_adapter_multi_1.enable_multi("first")
self.nest_adapter_multi_1b = self.scaffold.create_adapter(
"test_single_neuron")
self.nest_adapter_multi_1b.enable_multi("first")
self.nest_adapter_multi_2 = self.scaffold.create_adapter(
"test_single_neuron")
self.nest_adapter_multi_2.enable_multi("second")
def tearDown(self):
# Clean up any remaining locks to keep the test functions independent.
# Otherwise a chain reaction of failures is evoked.
self.nest_adapter_0.delete_lock()
def test_single_instance_unwanted_usage(self):
# Test AdapterError when trying to unlock unlocked adapter
self.assertRaises(AdapterError, self.nest_adapter_0.release_lock)
# Test whether the scaffold throws an AdapterError when the same
# adapter is prepared twice.
self.nest_adapter_0.prepare()
self.assertRaises(AdapterError, self.nest_adapter_0.prepare)
self.nest_adapter_0.release_lock()
self.nest_adapter_0.reset()
def test_single_instance_single_lock(self):
self.nest_adapter_1.reset()
# Lock kernel. Prepare adapter and thereby lock NEST kernel
self.nest_adapter_1.prepare()
lock_data = self.nest_adapter_1.read_lock()
self.assertEqual(lock_data["multi"], False)
self.assertEqual(self.nest_adapter_1.multi, False)
self.assertEqual(self.nest_adapter_1.has_lock, True)
# Release lock.
self.nest_adapter_1.release_lock()
self.assertEqual(self.nest_adapter_1.read_lock(), None)
self.assertEqual(self.nest_adapter_1.has_lock, False)
self.nest_adapter_1.reset()
def test_multi_instance_single_lock(self):
# Test that a 2nd single-instance adapter can't manage a locked kernel.
self.nest_adapter_1.prepare()
self.assertRaises(KernelLockedError, self.nest_adapter_2.prepare)
self.assertEqual(self.nest_adapter_2.is_prepared, False)
self.nest_adapter_1.release_lock()
self.nest_adapter_1.reset()
self.nest_adapter_2.reset()
def test_single_instance_multi_lock(self):
self.nest_adapter_multi_1.reset()
# Test functionality of the multi lock.
self.nest_adapter_multi_1.prepare()
lock_data = self.nest_adapter_multi_1.read_lock()
self.assertEqual(self.nest_adapter_multi_1.suffix, "first")
self.assertEqual(lock_data["multi"], True)
self.assertEqual(lock_data["suffixes"][0],
self.nest_adapter_multi_1.suffix)
self.assertEqual(self.nest_adapter_multi_1.multi, True)
self.assertEqual(self.nest_adapter_multi_1.is_prepared, True)
self.assertEqual(self.nest_adapter_multi_1.has_lock, True)
self.nest_adapter_multi_1.release_lock()
self.assertEqual(self.nest_adapter_multi_1.multi, True)
self.assertEqual(self.nest_adapter_multi_1.has_lock, False)
self.nest_adapter_multi_1.reset()
self.assertEqual(self.nest_adapter_multi_1.is_prepared, False)
def test_multi_instance_multi_lock(self):
# Test functionality of the multi lock.
self.nest_adapter_multi_1.prepare()
# Test that we have 1 lock.
lock_data = self.nest_adapter_multi_1.read_lock()
# Check multi instance multi lock
self.nest_adapter_multi_2.cell_models["test_cell"].parameters[
"t_ref"] = 3.0
self.nest_adapter_multi_2.prepare()
# Check that we have 2 locks
lock_data = self.nest_adapter_multi_1.read_lock()
self.assertEqual(len(lock_data["suffixes"]), 2)
# Test that we set the right parameters on the right classes.
try:
params = self.nest.GetDefaults("test_cell_first")
except Exception as e:
self.fail("First suffixed NEST models not found")
try:
params = self.nest.GetDefaults("test_cell_second")
except Exception as e:
self.fail("Second suffixed NEST models not found")
# Test that the adapters have the correct nest_identifiers
id1 = self.nest_adapter_multi_1.cell_models[
"test_cell"].nest_identifiers
id2 = self.nest_adapter_multi_2.cell_models[
"test_cell"].nest_identifiers
self.assertEqual(id1, [1, 2, 3, 4])
self.assertEqual(id2, [5, 6, 7, 8])
# Test that the adapter nodes have the right model
self.assertTrue(
all(
map(
lambda x: str(x["model"]) == "test_cell_first",
self.nest.GetStatus(id1),
)))
self.assertTrue(
all(
map(
lambda x: str(x["model"]) == "test_cell_second",
self.nest.GetStatus(id2),
)))
# Test that the adapter nodes have the right differential parameter t_ref
self.assertTrue(
all(map(lambda x: x["t_ref"] == 1.5, self.nest.GetStatus(id1))))
self.assertTrue(
all(map(lambda x: x["t_ref"] == 3.0, self.nest.GetStatus(id2))))
# Check duplicate suffixes
self.assertRaises(SuffixTakenError, self.nest_adapter_multi_1b.prepare)
self.nest_adapter_multi_1.release_lock()
self.nest_adapter_multi_1.reset()
# Test that we have 1 lock again after release.
lock_data = self.nest_adapter_multi_1.read_lock()
self.assertEqual(lock_data["suffixes"][0], "second")
self.nest_adapter_multi_2.release_lock()
self.nest_adapter_multi_2.reset()
class TestPlacementSets(unittest.TestCase):
"""
Check if the scaffold can create a single cell type.
"""
@classmethod
def setUpClass(self):
super(TestPlacementSets, self).setUpClass()
test_setup.prep_morphologies()
config = JSONConfig(file=double_neuron_config)
self.scaffold = Scaffold(config)
self.scaffold.compile_network()
def test_hdf5_structure(self):
with h5py.File(self.scaffold.output_formatter.file, "r") as h:
for key in ["from", "to"]:
group = h["cells/placement/" + key + "_cell"]
self.assertTrue(
"identifiers" in group,
"Identifiers dataset missing for the " + key + "_cell",
)
self.assertTrue(
"positions" in group,
"Positions dataset missing for the " + key + "_cell",
)
self.assertEqual(
group["positions"].shape,
(4, 3),
"Incorrect position dataset size for the " + key + "_cell",
)
self.assertTrue(
group["positions"].dtype == np.float64,
"Incorrect position dataset dtype ({}) for the ".format(
group["positions"].dtype) + key + "_cell",
)
self.assertEqual(
group["identifiers"].shape,
(2, ),
"Incorrect or noncontinuous identifiers dataset size for the "
+ key + "_cell",
)
self.assertTrue(
group["identifiers"].dtype == np.int32,
"Incorrect identifiers dataset dtype ({}) for the ".format(
group["identifiers"].dtype) + key + "_cell",
)
def test_placement_set_properties(self):
for key in ["from", "to"]:
cell_type = self.scaffold.get_cell_type(key + "_cell")
ps = PlacementSet(self.scaffold.output_formatter, cell_type)
self.assertEqual(
ps.identifiers.shape,
(4, ),
"Incorrect identifiers shape for " + key + "_cell",
)
self.assertEqual(
ps.positions.shape,
(4, 3),
"Incorrect positions shape for " + key + "_cell",
)
self.assertRaises(DatasetNotFoundError, lambda: ps.rotations)
self.assertEqual(type(ps.cells[0]), Cell,
"PlacementSet.cells did not return Cells")
self.assertEqual(
ps.cells[1].id,
1 if key == "from" else 5,
"PlacementSet.cells identifiers incorrect",
)
self.assertEqual(
ps.cells[1].position.shape,
(3, ),
"PlacementSet.cells positions wrong shape",
)
class TestFiberIntersection(unittest.TestCase):
@classmethod
def setUpClass(self):
super(TestFiberIntersection, self).setUpClass()
# Make sure the MR exists
test_setup.prep_morphologies()
# The scaffold has only the Granular layer (100x100x150) with 20 GrCs
# and 1 GoC placed, as specified in the config file
self.config = JSONConfig(file=fiber_transform_config)
# Defining quivers field to include also voxels outside the scaffold
# volume
self.quivers_field = np.zeros(
shape=(3, 80, 80, 80)) # Each voxel in the volume has vol_res=25um
# Fake quiver, oriented as original fibers
basic_quiver = np.array([0, 1.0, 0.0])
self.quivers_field[0, :] = basic_quiver[0]
self.quivers_field[1, :] = basic_quiver[1]
self.quivers_field[2, :] = basic_quiver[2]
self.config.connection_types[
"parallel_fiber_to_golgi_bended"].transformation.quivers = self.quivers_field
self.config.connection_types[
"parallel_fiber_to_golgi_bended"].transformation.vol_start = [
-500.0, -500.0, -500.0
]
self.scaffold = Scaffold(self.config)
self.scaffold.morphology_repository = MorphologyRepository(morpho_file)
self.scaffold.compile_network()
def test_fiber_connections(self):
pre_type = "granule_cell"
pre_neu = self.scaffold.get_placement_set(pre_type)
conn_type = "parallel_fiber_to_golgi"
cs = self.scaffold.get_connectivity_set(conn_type)
num_conn = len(cs.connections)
# Check that no more connections are formed than the number of
# presynaptic neurons - how could happen otherwise?
self.assertTrue(num_conn <= len(pre_neu.identifiers))
# Check that increasing resolution in FiberIntersection does not change
# connection number if there are no transformations (and thus the fibers
# are parallel to main axes)
conn_type_HR = "parallel_fiber_to_golgi_HR"
cs_HR = self.scaffold.get_connectivity_set(conn_type_HR)
self.assertEqual(num_conn, len(cs_HR.connections))
# Check that half (+- 5) connections are obtained with half the affinity
conn_type_affinity = "parallel_fiber_to_golgi_affinity"
cs_affinity = self.scaffold.get_connectivity_set(conn_type_affinity)
self.assertAlmostEqual(num_conn / 2,
len(cs_affinity.connections),
delta=5)
# Check that same number of connections are obtained when a fake quiver
# transformation is applied
conn_type_transform = "parallel_fiber_to_golgi_bended"
cs_fake_transform = self.scaffold.get_connectivity_set(
conn_type_transform)
self.assertEqual(len(cs_fake_transform.connections), num_conn)
# Check that less connections are obtained when the PC surface is
# oriented according to orientation vector of 45° rotation in yz plane,
# for how the Golgi cell is placed and the parallel fibers are rotated
basic_quiver = np.array([0, 0.7, 0.7])
self.quivers_field[0, :] = basic_quiver[0]
self.quivers_field[1, :] = basic_quiver[1]
self.quivers_field[2, :] = basic_quiver[2]
self.config.connection_types[
"parallel_fiber_to_golgi_bended"].transformation.quivers = self.quivers_field
self.scaffold = Scaffold(self.config)
self.scaffold.compile_network()
cs_transform = self.scaffold.get_connectivity_set(conn_type_transform)
self.assertTrue(len(cs_transform.connections) <= num_conn)