def test_hierarchical(self):
# most of leiden is tested in unit / integration tests in graspologic-native.
# All we're trying to test through these unit tests are the python conversions
# prior to calling, so type and value validation and that we got a result
edges = _create_edge_list()
results = hierarchical_leiden(edges, random_seed=1234)
total_nodes = len([item for item in results if item.level == 0])
partitions = HierarchicalCluster.final_hierarchical_clustering(results)
self.assertEqual(total_nodes, len(partitions))
def test_isolate_nodes_in_nx_graph_are_not_returned(self):
self.assertEqual(
10,
len(self.graph.nodes),
"the input graph contains all nodes including isolate",
)
with pytest.warns(UserWarning, match="isolate"):
partitions = leiden(self.graph)
self.assert_isolate_not_in_result(partitions)
with pytest.warns(UserWarning, match="isolate"):
hierarchical_partitions = hierarchical_leiden(self.graph)
self.assert_isolate_not_in_hierarchical_result(hierarchical_partitions)
def test_isolate_nodes_in_csr_matrix_are_not_returned(self):
sparse_adj_matrix = nx.to_scipy_sparse_matrix(self.graph)
self.assertEqual(
10,
sparse_adj_matrix.shape[0],
"the input csr contains all nodes including isolate",
)
with pytest.warns(UserWarning, match="isolate"):
partitions = leiden(sparse_adj_matrix)
self.assert_isolate_not_in_result(partitions)
with pytest.warns(UserWarning, match="isolate"):
hierarchical_partitions = hierarchical_leiden(sparse_adj_matrix)
self.assert_isolate_not_in_hierarchical_result(hierarchical_partitions)
def test_isolate_nodes_in_ndarray_are_not_returned(self):
ndarray_adj_matrix = nx.to_numpy_array(self.graph)
self.assertEqual(
10,
ndarray_adj_matrix.shape[0],
"the input array contains all nodes including isolate",
)
with pytest.warns(UserWarning, match="isolate"):
partitions = leiden(ndarray_adj_matrix)
self.assert_isolate_not_in_result(partitions)
with pytest.warns(UserWarning, match="isolate"):
hierarchical_partitions = hierarchical_leiden(ndarray_adj_matrix)
self.assert_isolate_not_in_hierarchical_result(hierarchical_partitions)
def test_correct_types(self):
# both leiden and hierarchical_leiden require the same types and mostly the same value range restrictions
good_args = {
"starting_communities": {
"1": 2
},
"extra_forced_iterations": 0,
"resolution": 1.0,
"randomness": 0.001,
"use_modularity": True,
"random_seed": None,
"is_weighted": True,
"weight_default": 1.0,
"check_directed": True,
}
graph = nx.Graph()
graph.add_edge("1", "2", weight=3.0)
graph.add_edge("2", "3", weight=4.0)
leiden(graph=graph, **good_args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["starting_communities"] = 123
leiden(graph=graph, **args)
args = good_args.copy()
args["starting_communities"] = None
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["extra_forced_iterations"] = 1234.003
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["extra_forced_iterations"] = -4003
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["resolution"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["resolution"] = 0
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["randomness"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["randomness"] = 0
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["use_modularity"] = 1234
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["trials"] = "hotdog"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["trials"] = 0
leiden(graph=graph, **args)
args = good_args.copy()
args["random_seed"] = 1234
leiden(graph=graph, **args)
args["random_seed"] = None
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["random_seed"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["random_seed"] = -1
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["is_weighted"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["weight_default"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["check_directed"] = "leiden"
leiden(graph=graph, **args)
# one extra parameter hierarchical needs
with self.assertRaises(TypeError):
args = good_args.copy()
args["max_cluster_size"] = "leiden"
hierarchical_leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["max_cluster_size"] = 0
hierarchical_leiden(graph=graph, **args)
as_csr = nx.to_scipy_sparse_matrix(graph)
partitions = leiden(graph=as_csr, **good_args)
node_ids = partitions.keys()
for node_id in node_ids:
self.assertTrue(
isinstance(node_id, (np.int32, np.intc)),
f"{node_id} has {type(node_id)} should be an np.int32/np.intc",
)
def test_correct_types(self):
# both leiden and hierarchical_leiden require the same types and mostly the same value range restrictions
good_args = {
"starting_communities": {"1": 2},
"extra_forced_iterations": 0,
"resolution": 1.0,
"randomness": 0.001,
"use_modularity": True,
"random_seed": None,
"is_weighted": True,
"weight_default": 1.0,
"check_directed": True,
}
graph = nx.Graph()
graph.add_edge("1", "2", weight=3.0)
graph.add_edge("2", "3", weight=4.0)
leiden(graph=graph, **good_args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["starting_communities"] = 123
leiden(graph=graph, **args)
args = good_args.copy()
args["starting_communities"] = None
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["extra_forced_iterations"] = 1234.003
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["extra_forced_iterations"] = -4003
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["resolution"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["resolution"] = 0
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["randomness"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["randomness"] = 0
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["use_modularity"] = 1234
leiden(graph=graph, **args)
args = good_args.copy()
args["random_seed"] = 1234
leiden(graph=graph, **args)
args["random_seed"] = None
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["random_seed"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["random_seed"] = -1
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["is_weighted"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["weight_default"] = "leiden"
leiden(graph=graph, **args)
with self.assertRaises(TypeError):
args = good_args.copy()
args["check_directed"] = "leiden"
leiden(graph=graph, **args)
# one extra parameter hierarchical needs
with self.assertRaises(TypeError):
args = good_args.copy()
args["max_cluster_size"] = "leiden"
hierarchical_leiden(graph=graph, **args)
with self.assertRaises(ValueError):
args = good_args.copy()
args["max_cluster_size"] = 0
hierarchical_leiden(graph=graph, **args)