def test_gframe(self):
g = SGraph()
v = g.vertices
self.assertSequenceEqual(v.column_names(), ['__id'])
e = g.edges
self.assertSequenceEqual(e.column_names(), ['__src_id', '__dst_id'])
# Test vertices and edge attributes cannot be modified
def set_vertices_empty(g):
g.vertices = SFrame()
def set_edges_empty(g):
g.edges = SFrame()
def remove_vertices(g):
del g.vertices
def remove_edges(g):
del g.edges
def remove_edge_column(gf, name):
del gf[name]
self.assertRaises(AttributeError, lambda: remove_vertices(g))
self.assertRaises(AttributeError, lambda: remove_edges(g))
self.assertRaises(AttributeError, lambda: set_vertices_empty(g))
self.assertRaises(AttributeError, lambda: set_edges_empty(g))
# Test gframe operations has the same effect as its sframe+graph equivalent
g = SGraph().add_vertices(self.vertices, 'vid').add_edges(self.edges, 'src_id', 'dst_id')
v = g.vertices
v['id_col'] = v['__id']
e = g.edges
e['src_id_col'] = e['__src_id']
e['dst_id_col'] = e['__dst_id']
g2 = SGraph().add_vertices(self.vertices, 'vid').add_edges(self.edges, 'src_id', 'dst_id')
new_vdata = g2.get_vertices()
new_vdata['id_col'] = new_vdata['__id']
new_edata = g2.get_edges()
new_edata['src_id_col'] = new_edata['__src_id']
new_edata['dst_id_col'] = new_edata['__dst_id']
g2 = SGraph().add_vertices(new_vdata, '__id').add_edges(new_edata, '__src_id', '__dst_id')
assert_frame_equal(g.get_vertices().to_dataframe().sort('__id').reset_index(drop=True),
g2.get_vertices().to_dataframe().sort('__id').reset_index(drop=True))
assert_frame_equal(g.get_edges().to_dataframe().sort(['__src_id', '__dst_id']).reset_index(drop=True),
g2.get_edges().to_dataframe().sort(['__src_id', '__dst_id']).reset_index(drop=True))
# check delete a column with exception, and edges is still in a valid state
self.assertRaises(KeyError, lambda: remove_edge_column(g.edges, 'badcolumn'))
g.edges.head()
# test slicing
assert_frame_equal(g.edges[:3].to_dataframe(), g.get_edges()[:3].to_dataframe())
assert_frame_equal(g.vertices[:3].to_dataframe(), g.get_vertices()[:3].to_dataframe())
# test add row number
e_expected = g.get_edges().to_dataframe()
v_expected = g.get_vertices().to_dataframe()
e_expected['id'] = range(len(e_expected))
v_expected['id'] = range(len(v_expected))
def test_empty_graph(self):
g = SGraph()
self.assertEqual(g.summary(), {'num_vertices': 0, 'num_edges': 0})
self.assertEqual(len(g.get_fields()), 3)
self.assertTrue(g.get_vertices(format='sframe').shape, (0, 1))
self.assertTrue(g.get_edges(format='sframe').shape, (0, 2))
self.assertTrue(g.vertices.shape, (0, 1))
self.assertTrue(g.edges.shape, (0, 2))
self.assertTrue(len(g.get_vertices(format='list')) == 0)
self.assertTrue(len(g.get_edges(format='list')) == 0)
def test_empty_graph(self):
g = SGraph()
self.assertEqual(g.summary(), {'num_vertices': 0, 'num_edges': 0})
self.assertEqual(len(g.get_fields()), 3)
self.assertTrue(g.get_vertices(format='sframe').shape, (0, 1))
self.assertTrue(g.get_edges(format='sframe').shape, (0, 2))
self.assertTrue(g.vertices.shape, (0, 1))
self.assertTrue(g.edges.shape, (0, 2))
self.assertTrue(len(g.get_vertices(format='list')) == 0)
self.assertTrue(len(g.get_edges(format='list')) == 0)
def test_simple_graph(self):
for input_type in [pd.DataFrame, SFrame, list]:
g = SGraph()
if input_type is list:
vertices = [Vertex(x[1]['vid'], {'color': x[1]['color'], 'vec': x[1]['vec']}) for x in self.vertices.iterrows()]
edges = [Edge(x[1]['src_id'], x[1]['dst_id'], {'weight': x[1]['weight']}) for x in self.edges.iterrows()]
g = g.add_vertices(vertices)
g = g.add_edges(edges)
else:
g = g.add_vertices(input_type(self.vertices), vid_field='vid')
g = g.add_edges(input_type(self.edges), src_field='src_id', dst_field='dst_id')
self.assertEqual(g.summary(), {'num_vertices': 4, 'num_edges': 3})
self.assertItemsEqual(g.get_fields(), ['__id', '__src_id', '__dst_id', 'color', 'vec', 'weight'])
self.assertItemsEqual(g.get_vertices(format='dataframe').columns.values, ['color', 'vec'])
self.assertItemsEqual(g.get_edges(format='dataframe').columns.values, ['__src_id', '__dst_id', 'weight'])
self.assertTrue(g.get_edges(format='dataframe').shape, (3, 3))
self.assertTrue(g.get_vertices(format='dataframe').shape, (4, 3))
self.assertTrue(g.get_vertices(format='dataframe', fields={'color': 'g'}).shape, (1, 2))
self.assertTrue(g.get_edges(format='dataframe', fields={'weight': 0.}).shape, (1, 3))
self.assertItemsEqual(g.get_vertices(format='sframe').column_names(), ['__id', 'color', 'vec'])
self.assertItemsEqual(g.get_edges(format='sframe').column_names(), ['__src_id', '__dst_id', 'weight'])
self.assertTrue(g.get_edges(format='sframe').shape, (3, 3))
self.assertTrue(g.get_vertices(format='sframe').shape, (4, 3))
self.assertTrue(g.get_vertices(format='sframe', fields={'color': 'g'}).shape, (1, 2))
self.assertTrue(g.get_edges(format='sframe', fields={'weight': 0.}).shape, (1, 3))
vertices = g.get_vertices(format='list')
edges = g.get_edges(format='list')
self.assertEqual(len(vertices), 4)
self.assertEqual(len(edges), 3)
# get edges is lazy
edges = g.get_edges()
self.assertFalse(edges.__is_materialized__())
def test_robust_parse(self):
df = pd.DataFrame({'int': [1, 2, 3],
'float': [1., 2., 3.],
'str': ['one', 'two', 'three'],
'nan': [np.nan, np.nan, np.nan],
'sparse_int': [1, 2, np.nan],
'sparse_float': [np.nan, 2., 3.],
'sparse_str': [None, 'two', None]
})
g = SGraph().add_vertices(df)
self.assertItemsEqual(g.get_fields(), df.columns.tolist() + ['__id', '__src_id', '__dst_id'])
df2 = g.get_vertices(format='dataframe')
sf = g.get_vertices(format='sframe')
for col in df.columns:
# potential bug: df2 is missing the 'nan' column.
if (col != 'nan'):
self.assertItemsEqual(sorted(list(df2[col].dropna())), sorted(list(df[col].dropna())))
self.assertItemsEqual(sorted(list(sf[col].dropna())), sorted(list(df[col].dropna())))
def test_robust_parse(self):
df = pd.DataFrame({
'int': [1, 2, 3],
'float': [1., 2., 3.],
'str': ['one', 'two', 'three'],
'nan': [np.nan, np.nan, np.nan],
'sparse_int': [1, 2, np.nan],
'sparse_float': [np.nan, 2., 3.],
'sparse_str': [None, 'two', None]
})
g = SGraph().add_vertices(df)
self.assertItemsEqual(
g.get_fields(),
df.columns.tolist() + ['__id', '__src_id', '__dst_id'])
df2 = g.get_vertices(format='dataframe')
sf = g.get_vertices(format='sframe')
for col in df.columns:
# potential bug: df2 is missing the 'nan' column.
if (col != 'nan'):
self.assertItemsEqual(sorted(list(df2[col].dropna())),
sorted(list(df[col].dropna())))
self.assertItemsEqual(sorted(list(sf[col].dropna())),
sorted(list(df[col].dropna())))
def test_simple_graph(self):
for input_type in [pd.DataFrame, SFrame, list]:
g = SGraph()
if input_type is list:
vertices = [
Vertex(x[1]['vid'], {
'color': x[1]['color'],
'vec': x[1]['vec']
}) for x in self.vertices.iterrows()
]
edges = [
Edge(x[1]['src_id'], x[1]['dst_id'],
{'weight': x[1]['weight']})
for x in self.edges.iterrows()
]
g = g.add_vertices(vertices)
g = g.add_edges(edges)
else:
g = g.add_vertices(input_type(self.vertices), vid_field='vid')
g = g.add_edges(input_type(self.edges),
src_field='src_id',
dst_field='dst_id')
self.assertEqual(g.summary(), {'num_vertices': 4, 'num_edges': 3})
self.assertItemsEqual(
g.get_fields(),
['__id', '__src_id', '__dst_id', 'color', 'vec', 'weight'])
self.assertItemsEqual(
g.get_vertices(format='dataframe').columns.values,
['color', 'vec'])
self.assertItemsEqual(
g.get_edges(format='dataframe').columns.values,
['__src_id', '__dst_id', 'weight'])
self.assertTrue(g.get_edges(format='dataframe').shape, (3, 3))
self.assertTrue(g.get_vertices(format='dataframe').shape, (4, 3))
self.assertTrue(
g.get_vertices(format='dataframe', fields={
'color': 'g'
}).shape, (1, 2))
self.assertTrue(
g.get_edges(format='dataframe', fields={
'weight': 0.
}).shape, (1, 3))
self.assertItemsEqual(
g.get_vertices(format='sframe').column_names(),
['__id', 'color', 'vec'])
self.assertItemsEqual(
g.get_edges(format='sframe').column_names(),
['__src_id', '__dst_id', 'weight'])
self.assertTrue(g.get_edges(format='sframe').shape, (3, 3))
self.assertTrue(g.get_vertices(format='sframe').shape, (4, 3))
self.assertTrue(
g.get_vertices(format='sframe', fields={
'color': 'g'
}).shape, (1, 2))
self.assertTrue(
g.get_edges(format='sframe', fields={
'weight': 0.
}).shape, (1, 3))
vertices = g.get_vertices(format='list')
edges = g.get_edges(format='list')
self.assertEqual(len(vertices), 4)
self.assertEqual(len(edges), 3)
# get edges is lazy
edges = g.get_edges()
self.assertFalse(edges.__is_materialized__())
def test_gframe(self):
g = SGraph()
v = g.vertices
self.assertSequenceEqual(v.column_names(), ['__id'])
e = g.edges
self.assertSequenceEqual(e.column_names(), ['__src_id', '__dst_id'])
# Test vertices and edge attributes cannot be modified
def set_vertices_empty(g):
g.vertices = SFrame()
def set_edges_empty(g):
g.edges = SFrame()
def remove_vertices(g):
del g.vertices
def remove_edges(g):
del g.edges
def remove_edge_column(gf, name):
del gf[name]
self.assertRaises(AttributeError, lambda: remove_vertices(g))
self.assertRaises(AttributeError, lambda: remove_edges(g))
self.assertRaises(AttributeError, lambda: set_vertices_empty(g))
self.assertRaises(AttributeError, lambda: set_edges_empty(g))
# Test gframe operations has the same effect as its sframe+graph equivalent
g = SGraph().add_vertices(self.vertices,
'vid').add_edges(self.edges, 'src_id',
'dst_id')
v = g.vertices
v['id_col'] = v['__id']
e = g.edges
e['src_id_col'] = e['__src_id']
e['dst_id_col'] = e['__dst_id']
g2 = SGraph().add_vertices(self.vertices,
'vid').add_edges(self.edges, 'src_id',
'dst_id')
new_vdata = g2.get_vertices()
new_vdata['id_col'] = new_vdata['__id']
new_edata = g2.get_edges()
new_edata['src_id_col'] = new_edata['__src_id']
new_edata['dst_id_col'] = new_edata['__dst_id']
g2 = SGraph().add_vertices(new_vdata,
'__id').add_edges(new_edata, '__src_id',
'__dst_id')
assert_frame_equal(
g.get_vertices().to_dataframe().sort('__id').reset_index(
drop=True),
g2.get_vertices().to_dataframe().sort('__id').reset_index(
drop=True))
assert_frame_equal(
g.get_edges().to_dataframe().sort(['__src_id', '__dst_id'
]).reset_index(drop=True),
g2.get_edges().to_dataframe().sort(['__src_id', '__dst_id'
]).reset_index(drop=True))
# check delete a column with exception, and edges is still in a valid state
self.assertRaises(KeyError,
lambda: remove_edge_column(g.edges, 'badcolumn'))
g.edges.head()
# test slicing
assert_frame_equal(g.edges[:3].to_dataframe(),
g.get_edges()[:3].to_dataframe())
assert_frame_equal(g.vertices[:3].to_dataframe(),
g.get_vertices()[:3].to_dataframe())
# test add row number
e_expected = g.get_edges().to_dataframe()
v_expected = g.get_vertices().to_dataframe()
e_expected['id'] = range(len(e_expected))
v_expected['id'] = range(len(v_expected))
