def test_df_to_tsv(self):
path = os.path.join(tempfile.mkdtemp(), 'some.tsv')
df = tsv_to_df(self.edges_file)
df_to_tsv(df, path)
self.assertTrue(os.path.isfile(path))
df_roundtrip = tsv_to_df(path)
self.assertEqual(df.shape, df_roundtrip.shape)
def test_has_disconnected_nodes(self):
nodes_extra_ids = tsv_to_df(
'tests/resources/edges/bigger_graph_nodes_EXTRA_IDS.tsv')
nodes_missing_ids = tsv_to_df(
'tests/resources/edges/bigger_graph_nodes_MISSING_IDS.tsv')
self.assertTrue(not has_disconnected_nodes(edges_df=self.edges,
nodes_df=self.nodes))
with self.assertWarns(Warning):
self.assertTrue(not has_disconnected_nodes(edges_df=self.edges,
nodes_df=nodes_missing_ids))
self.assertTrue(has_disconnected_nodes(edges_df=self.edges,
nodes_df=nodes_extra_ids))
def setUpClass(cls) -> None:
cls.nodes_file = 'tests/resources/edges/bigger_graph_nodes.tsv'
cls.edges_file = 'tests/resources/edges/bigger_graph_edges.tsv'
cls.edges = tsv_to_df(cls.edges_file)
cls.nodes = tsv_to_df(cls.nodes_file)
# make negative edges for small graph
cls.ne = make_negative_edges(nodes_df=cls.nodes, edges_df=cls.edges)
# make positive edges for small graph
cls.train_fraction = 0.8
(cls.train_edges, cls.test_edges) = make_positive_edges(
nodes_df=cls.nodes, edges_df=cls.edges, train_fraction= cls.train_fraction,
min_degree=0)
def test_make_edges_check_node_output_file(self):
output_dir = tempfile.mkdtemp()
output_file_with_path = os.path.join(output_dir, 'pos_train_nodes.tsv')
input_nodes = tsv_to_df(self.nodes_file)
make_edges(nodes=self.nodes_file, edges=self.edges_file,
output_dir=output_dir, train_fraction=0.8,
validation=False, min_degree=1)
self.assertTrue(os.path.isfile(output_file_with_path))
new_nodes_df = tsv_to_df(output_file_with_path)
# make sure we get expected
self.assertAlmostEqual(new_nodes_df.shape[0], input_nodes.shape[0])
# should also have subject and object column
self.assertTrue('id' in new_nodes_df)
self.assertTrue('category' in new_nodes_df)
def test_make_edges_check_edge_output_files(self, output_file: str,
make_validation: bool,
file_should_exist: bool,
expected_fract: float):
me_output_dir = tempfile.mkdtemp()
output_file_with_path = os.path.join(me_output_dir, output_file)
input_edges = tsv_to_df(self.edges_file)
num_input_edges = input_edges.shape[0]
make_edges(nodes=self.nodes_file, edges=self.edges_file,
output_dir=me_output_dir, train_fraction=0.8,
validation=make_validation, min_degree=1)
if file_should_exist:
self.assertTrue(os.path.isfile(output_file_with_path))
new_edges_df = tsv_to_df(output_file_with_path)
# make sure we get expected
self.assertAlmostEqual(new_edges_df.shape[0],
num_input_edges * expected_fract, 1)
# should also have subject and object column
self.assertTrue('subject' in new_edges_df)
self.assertTrue('object' in new_edges_df)
else:
self.assertTrue(not os.path.isfile(output_file_with_path))
def test_make_edges_pos_train_test_valid_edges_distinct(self, train, test, valid):
output_dir = tempfile.mkdtemp()
input_edges = tsv_to_df(self.edges_file)
make_edges(nodes=self.nodes_file, edges=self.edges_file,
output_dir=output_dir, train_fraction=0.8,
validation=True, min_degree=1)
input_edges = tsv_to_df(self.edges_file)[['subject', 'object']]
train_edges = tsv_to_df(os.path.join(output_dir, train))[['subject', 'object']]
test_edges = tsv_to_df(os.path.join(output_dir, test))[['subject', 'object']]
valid_edges = tsv_to_df(os.path.join(output_dir, valid))[['subject', 'object']]
# train should not share any members with test
self.assertTrue(not set(train_edges).isdisjoint(test_edges))
# train should not share any members with valid
self.assertTrue(not set(train_edges).isdisjoint(valid_edges))
# test should not share any members with valid
self.assertTrue(not set(test_edges).isdisjoint(valid_edges))
# train should be a subset of input_edges
self.assertTrue(set(train_edges) <= set(input_edges))
# test should be a subset of input_edges
self.assertTrue(set(test_edges) <= set(input_edges))
# valid should be a subset of input_edges
self.assertTrue(set(valid_edges) <= set(input_edges))
def test_make_positive_edges_test_min_degree_gt_zero(self):
train_fraction = 0.90
degree = 2
hd_edges_file =\
'tests/resources/edges/bigger_graph_edges_HIGHER_DEGREE_NODES.tsv'
hd_edges = tsv_to_df(hd_edges_file)
hd_nodes = ['p1', 'd1',
'g1', 'g2', 'g3', 'g4', 'g5', 'g6', 'g7', 'g8', 'g9', 'g10',
'g11', 'g12', 'g13', 'g14', 'g15', 'g16', 'g17', 'g18', 'g19',
'g20', 'g21', 'g22', 'g23', 'g24', 'g25']
for _ in range(10):
(train_edges, test_edges) = make_positive_edges(
nodes_df=self.nodes, edges_df=hd_edges, train_fraction=train_fraction,
min_degree=degree)
these_nodes = set(list(test_edges.subject) + list(test_edges.object))
self.assertTrue(set(these_nodes) < set(hd_nodes),
"Got some nodes with degree < 2: %s" %
" ".join(np.setdiff1d(these_nodes,hd_nodes)[0]))
def test_tsv_to_df(self):
df = tsv_to_df(self.edges_file)
self.assertTrue(isinstance(df, pd.DataFrame))
self.assertEqual((150, 5), df.shape)
self.assertEqual(df['subject'][0], 'g1')