def test_add_edge_data_prop_columns(df_type):
"""
add_edge_data() on "transactions" table, subset of properties.
"""
from cugraph.experimental import PropertyGraph
transactions = dataset1["transactions"]
transactions_df = df_type(columns=transactions[0], data=transactions[1])
expected_props = ["card_num", "card_type"]
pG = PropertyGraph()
pG.add_edge_data(transactions_df,
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns=expected_props)
assert pG.num_vertices == 7
assert pG.num_edges == 4
assert sorted(pG.edge_property_names) == sorted(expected_props)
def test_add_vertex_data_prop_columns(df_type):
"""
add_vertex_data() on "merchants" table, subset of properties.
"""
from cugraph.experimental import PropertyGraph
merchants = dataset1["merchants"]
merchants_df = df_type(columns=merchants[0], data=merchants[1])
expected_props = ["merchant_name", "merchant_sales", "merchant_location"]
pG = PropertyGraph()
pG.add_vertex_data(merchants_df,
type_name="merchants",
vertex_id_column="merchant_id",
property_columns=expected_props)
assert pG.num_vertices == 5
assert pG.num_edges == 0
assert sorted(pG.vertex_property_names) == sorted(expected_props)
def test_add_vertex_data(df_type):
"""
add_vertex_data() on "merchants" table, all properties.
"""
from cugraph.experimental import PropertyGraph
merchants = dataset1["merchants"]
merchants_df = df_type(columns=merchants[0], data=merchants[1])
pG = PropertyGraph()
pG.add_vertex_data(merchants_df,
type_name="merchants",
vertex_id_column="merchant_id",
property_columns=None)
assert pG.num_vertices == 5
assert pG.num_edges == 0
expected_props = merchants[0].copy()
assert sorted(pG.vertex_property_names) == sorted(expected_props)
def test_null_data(df_type):
"""
test for null data
"""
from cugraph.experimental import PropertyGraph
pG = PropertyGraph()
assert pG.num_vertices == 0
assert pG.num_edges == 0
assert sorted(pG.vertex_property_names) == sorted([])
def read_reddit(raw_path, self_loop=False):
coo_adj = sp.load_npz(os.path.join(raw_path, "reddit_graph.npz"))
edgelist = cudf.DataFrame()
edgelist['src'] = cudf.Series(coo_adj.row)
edgelist['dst'] = cudf.Series(coo_adj.col)
edgelist['wt'] = cudf.Series(coo_adj.data)
# features and labels
reddit_data = np.load(os.path.join(raw_path, "reddit_data.npz"))
features = reddit_data["feature"]
cu_features = cudf.DataFrame(features)
cu_features['name'] = np.arange(cu_features.shape[0])
labels = reddit_data["label"]
# tarin/val/test indices
node_types = reddit_data["node_types"]
train_mask = (node_types == 1)
val_mask = (node_types == 2)
test_mask = (node_types == 3)
# add features to nodes and edges
pg = PropertyGraph()
pg.add_edge_data(edgelist, vertex_col_names=("src", "dst"))
pg.add_vertex_data(cu_features, vertex_col_name="name")
pg._vertex_prop_dataframe.drop(columns=['name'], inplace=True)
gstore = CuGraphStorage(pg)
return gstore, labels, train_mask, val_mask, test_mask
def read_cora(graph_path, feat_path, self_loop=False):
cora_M = cudf.read_csv(graph_path, sep='\t', header=None)
cora_content = cudf.read_csv(feat_path, sep='\t', header=None)
# the last column is true label
labels = cora_content['1434']
cora_content.drop(columns='1434', inplace=True)
# add weight into graph
cora_M['weight'] = 1.0
# add features to nodes and edges
pg = PropertyGraph()
pg.add_edge_data(cora_M, vertex_col_names=("0", "1"))
pg.add_vertex_data(cora_content, vertex_col_name="0")
pg._vertex_prop_dataframe.drop(columns=['0'], inplace=True)
pg._edge_prop_dataframe.drop(columns=['0', '1'], inplace=True)
gstore = CuGraphStorage(pg)
# define train, test and val splits
indices = np.arange(len(labels))
random.shuffle(indices)
idx_train, idx_val, idx_test = np.split(indices, [1000, 1500])
return gstore, labels, idx_train, idx_val, idx_test
def test_extract_subgraph_graph_without_vert_props():
"""
Ensure a subgraph can be extracted from a PropertyGraph that does not have
vertex properties.
"""
from cugraph.experimental import PropertyGraph
transactions = dataset1["transactions"]
relationships = dataset1["relationships"]
pG = PropertyGraph()
pG.add_edge_data(cudf.DataFrame(columns=transactions[0],
data=transactions[1]),
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns=None)
pG.add_edge_data(cudf.DataFrame(columns=relationships[0],
data=relationships[1]),
type_name="relationships",
vertex_id_columns=("user_id_1", "user_id_2"),
property_columns=None)
G = pG.extract_subgraph(selection=pG.select_edges("_SRC_ == 89216"),
create_using=DiGraph_inst,
edge_weight_property="relationship_type",
default_edge_weight=0)
expected_edgelist = cudf.DataFrame({
"src": [89216, 89216, 89216],
"dst": [4, 89021, 32431],
"weights": [0, 9, 9]
})
actual_edgelist = G.unrenumber(G.edgelist.edgelist_df,
"src",
preserve_order=True)
actual_edgelist = G.unrenumber(actual_edgelist, "dst", preserve_order=True)
assert G.is_directed()
assert_frame_equal(expected_edgelist, actual_edgelist, check_like=True)
def create_pg():
"""
Fixture which returns an instance of a PropertyGraph with vertex and edge
data added from dataset1, parameterized for different DataFrame types.
"""
dataframe_type = cudf.DataFrame
(merchants, users, taxpayers, transactions, relationships,
referrals) = dataset1.values()
pG = PropertyGraph()
# Vertex and edge data is added as one or more DataFrames; either a Pandas
# DataFrame to keep data on the CPU, a cuDF DataFrame to keep data on GPU,
# or a dask_cudf DataFrame to keep data on distributed GPUs.
# For dataset1: vertices are merchants and users, edges are transactions,
# relationships, and referrals.
# property_columns=None (the default) means all columns except
# vertex_col_name will be used as properties for the vertices/edges.
pG.add_vertex_data(dataframe_type(columns=merchants[0], data=merchants[1]),
type_name="merchants",
vertex_col_name="merchant_id",
property_columns=None)
pG.add_vertex_data(dataframe_type(columns=users[0], data=users[1]),
type_name="users",
vertex_col_name="user_id",
property_columns=None)
pG.add_vertex_data(dataframe_type(columns=taxpayers[0], data=taxpayers[1]),
type_name="taxpayers",
vertex_col_name="payer_id",
property_columns=None)
pG.add_edge_data(dataframe_type(columns=transactions[0],
data=transactions[1]),
type_name="transactions",
vertex_col_names=("user_id", "merchant_id"),
property_columns=None)
pG.add_edge_data(dataframe_type(columns=relationships[0],
data=relationships[1]),
type_name="relationships",
vertex_col_names=("user_id_1", "user_id_2"),
property_columns=None)
pG.add_edge_data(dataframe_type(columns=referrals[0], data=referrals[1]),
type_name="referrals",
vertex_col_names=("user_id_1", "user_id_2"),
property_columns=None)
return pG
def test_different_vertex_edge_input_dataframe_types():
"""
Ensures that a PropertyGraph initialized with one DataFrame type cannot be
extended with another.
"""
df = cudf.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
pdf = pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]})
from cugraph.experimental import PropertyGraph
pG = PropertyGraph()
pG.add_vertex_data(df, type_name="foo", vertex_id_column="a")
with pytest.raises(TypeError):
pG.add_edge_data(pdf, type_name="bar", vertex_id_columns=("a", "b"))
pG = PropertyGraph()
pG.add_vertex_data(pdf, type_name="foo", vertex_id_column="a")
with pytest.raises(TypeError):
pG.add_edge_data(df, type_name="bar", vertex_id_columns=("a", "b"))
# Different order
pG = PropertyGraph()
pG.add_edge_data(df, type_name="bar", vertex_id_columns=("a", "b"))
with pytest.raises(TypeError):
pG.add_vertex_data(pdf, type_name="foo", vertex_id_column="a")
# Same API call, different types
pG = PropertyGraph()
pG.add_vertex_data(df, type_name="foo", vertex_id_column="a")
with pytest.raises(TypeError):
pG.add_vertex_data(pdf, type_name="foo", vertex_id_column="a")
pG = PropertyGraph()
pG.add_edge_data(df, type_name="bar", vertex_id_columns=("a", "b"))
with pytest.raises(TypeError):
pG.add_edge_data(pdf, type_name="bar", vertex_id_columns=("a", "b"))
def test_add_edge_data_bad_args():
"""
add_edge_data() with various bad args, checks that proper exceptions are
raised.
"""
from cugraph.experimental import PropertyGraph
transactions = dataset1["transactions"]
transactions_df = cudf.DataFrame(columns=transactions[0],
data=transactions[1])
pG = PropertyGraph()
with pytest.raises(TypeError):
pG.add_edge_data(42,
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns=None)
with pytest.raises(TypeError):
pG.add_edge_data(transactions_df,
type_name=42,
vertex_id_columns=("user_id", "merchant_id"),
property_columns=None)
with pytest.raises(ValueError):
pG.add_edge_data(transactions_df,
type_name="transactions",
vertex_id_columns=("user_id", "bad_column"),
property_columns=None)
with pytest.raises(ValueError):
pG.add_edge_data(transactions_df,
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns=["bad_column_name", "time"])
with pytest.raises(TypeError):
pG.add_edge_data(transactions_df,
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns="time")
def test_add_vertex_data_bad_args():
"""
add_vertex_data() with various bad args, checks that proper exceptions are
raised.
"""
from cugraph.experimental import PropertyGraph
merchants = dataset1["merchants"]
merchants_df = cudf.DataFrame(columns=merchants[0], data=merchants[1])
pG = PropertyGraph()
with pytest.raises(TypeError):
pG.add_vertex_data(42,
type_name="merchants",
vertex_id_column="merchant_id",
property_columns=None)
with pytest.raises(TypeError):
pG.add_vertex_data(merchants_df,
type_name=42,
vertex_id_column="merchant_id",
property_columns=None)
with pytest.raises(ValueError):
pG.add_vertex_data(merchants_df,
type_name="merchants",
vertex_id_column="bad_column_name",
property_columns=None)
with pytest.raises(ValueError):
pG.add_vertex_data(
merchants_df,
type_name="merchants",
vertex_id_column="merchant_id",
property_columns=["bad_column_name", "merchant_name"])
with pytest.raises(TypeError):
pG.add_vertex_data(merchants_df,
type_name="merchants",
vertex_id_column="merchant_id",
property_columns="merchant_name")
def property_graph_instance(request):
"""
FIXME: fill this in
"""
dataframe_type = request.param[0]
from cugraph.experimental import PropertyGraph
(merchants, users, taxpayers, transactions, relationships,
referrals) = dataset1.values()
pG = PropertyGraph()
# Vertex and edge data is added as one or more DataFrames; either a Pandas
# DataFrame to keep data on the CPU, a cuDF DataFrame to keep data on GPU,
# or a dask_cudf DataFrame to keep data on distributed GPUs.
# For dataset1: vertices are merchants and users, edges are transactions,
# relationships, and referrals.
# property_columns=None (the default) means all columns except
# vertex_id_column will be used as properties for the vertices/edges.
pG.add_vertex_data(dataframe_type(columns=merchants[0], data=merchants[1]),
type_name="merchants",
vertex_id_column="merchant_id",
property_columns=None)
pG.add_vertex_data(dataframe_type(columns=users[0], data=users[1]),
type_name="users",
vertex_id_column="user_id",
property_columns=None)
pG.add_vertex_data(dataframe_type(columns=taxpayers[0], data=taxpayers[1]),
type_name="taxpayers",
vertex_id_column="payer_id",
property_columns=None)
pG.add_edge_data(dataframe_type(columns=transactions[0],
data=transactions[1]),
type_name="transactions",
vertex_id_columns=("user_id", "merchant_id"),
property_columns=None)
pG.add_edge_data(dataframe_type(columns=relationships[0],
data=relationships[1]),
type_name="relationships",
vertex_id_columns=("user_id_1", "user_id_2"),
property_columns=None)
pG.add_edge_data(dataframe_type(columns=referrals[0], data=referrals[1]),
type_name="referrals",
vertex_id_columns=("user_id_1", "user_id_2"),
property_columns=None)
return pG