Fixing Nx and Graph/DiGraph issues

python/cugraph/cugraph/centrality/betweenness_centrality.py

@@ -20,7 +20,6 @@
                                df_score_to_dictionary,
                                ensure_cugraph_obj_for_nx,
                                )
-import cugraph
 
 
 # NOTE: result_type=float could be an intuitive way to indicate the result type
@@ -262,7 +261,7 @@ def edge_betweenness_centrality(
         df = G.unrenumber(df, "src")
         df = G.unrenumber(df, "dst")
 
-    if type(G) is cugraph.Graph:
+    if G.is_directed() is False:
         # select the lower triangle of the df based on src/dst vertex value
         lower_triangle = df['src'] >= df['dst']
         # swap the src and dst vertices for the lower triangle only. Because

python/cugraph/cugraph/structure/graph_implementation/simpleDistributedGraph.py

@@ -88,7 +88,7 @@ def __from_edgelist(
                 raise Exception(
                     "edge_attr column name not found in input."
                     "Recheck the edge_attr parameter")
-            self.weighted = True
+            self.properties.weighted = True
             ddf_columns = ddf_columns + [edge_attr]
         input_ddf = input_ddf[ddf_columns]
 

python/cugraph/cugraph/structure/graph_implementation/simpleGraph.py

@@ -152,7 +152,7 @@ def __from_edgelist(
         dest_col = elist[destination]
 
         if edge_attr is not None:
-            self.weighted = True
+            self.properties.weighted = True
             value_col = elist[edge_attr]
         else:
             value_col = None

python/cugraph/cugraph/tests/test_bfs.py

@@ -21,6 +21,15 @@
 from cugraph.tests import utils
 import random
 
+import pandas as pd
+import cupy as cp
+from cupyx.scipy.sparse.coo import coo_matrix as cp_coo_matrix
+from cupyx.scipy.sparse.csr import csr_matrix as cp_csr_matrix
+from cupyx.scipy.sparse.csc import csc_matrix as cp_csc_matrix
+from scipy.sparse.coo import coo_matrix as sp_coo_matrix
+from scipy.sparse.csr import csr_matrix as sp_csr_matrix
+from scipy.sparse.csc import csc_matrix as sp_csc_matrix
+
 # Temporarily suppress warnings till networkX fixes deprecation warnings
 # (Using or importing the ABCs from 'collections' instead of from
 # 'collections.abc' is deprecated, and in 3.8 it will stop working) for
@@ -33,14 +42,6 @@
     import networkx as nx
     import networkx.algorithms.centrality.betweenness as nxacb
 
-import pandas as pd
-import cupy as cp
-from cupyx.scipy.sparse.coo import coo_matrix as cp_coo_matrix
-from cupyx.scipy.sparse.csr import csr_matrix as cp_csr_matrix
-from cupyx.scipy.sparse.csc import csc_matrix as cp_csc_matrix
-from scipy.sparse.coo import coo_matrix as sp_coo_matrix
-from scipy.sparse.csr import csr_matrix as sp_csr_matrix
-from scipy.sparse.csc import csc_matrix as sp_csc_matrix
 
 # =============================================================================
 # Parameters

python/cugraph/cugraph/tests/test_edge_betweenness_centrality.py

@@ -496,6 +496,7 @@ def test_edge_betweenness_centrality_nx(
     prepare_test()
 
     Gnx = utils.generate_nx_graph_from_file(graph_file, directed, edgevals)
+    assert nx.is_directed(Gnx) == directed
 
     nx_bc = nx.edge_betweenness_centrality(Gnx)
     cu_bc = cugraph.edge_betweenness_centrality(Gnx)
@@ -504,7 +505,8 @@ def test_edge_betweenness_centrality_nx(
     networkx_bc = sorted(nx_bc.items(), key=lambda x: x[0])
     cugraph_bc = sorted(cu_bc.items(), key=lambda x: x[0])
     err = 0
-    assert len(cugraph_bc) == len(networkx_bc)
+
+    assert len(networkx_bc) == len(cugraph_bc)
     for i in range(len(cugraph_bc)):
         if (
             abs(cugraph_bc[i][1] - networkx_bc[i][1]) > 0.01

python/cugraph/cugraph/tests/test_modularity.py

@@ -19,6 +19,9 @@
 import cudf
 import cugraph
 from cugraph.tests import utils
+from cugraph.utilities import ensure_cugraph_obj_for_nx
+
+import networkx as nx
 
 
 def cugraph_call(G, partitions):
@@ -51,7 +54,6 @@ def random_call(G, partitions):
 PARTITIONS = [2, 4, 8]
 
 
-# Test all combinations of default/managed and pooled/non-pooled allocation
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
 @pytest.mark.parametrize("partitions", PARTITIONS)
 def test_modularity_clustering(graph_file, partitions):
@@ -71,12 +73,38 @@ def test_modularity_clustering(graph_file, partitions):
     assert cu_score > rand_score
 
 
-# Test all combinations of default/managed and pooled/non-pooled allocation
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
 @pytest.mark.parametrize("partitions", PARTITIONS)
-def test_modularity_clustering_multi_column(graph_file, partitions):
-    gc.collect()
+def test_modularity_clustering_nx(graph_file, partitions):
+    # Read in the graph and get a cugraph object
+    csv_data = utils.read_csv_for_nx(graph_file, read_weights_in_sp=True)
+
+    nxG = nx.from_pandas_edgelist(
+            csv_data,
+            source="0",
+            target="1",
+            edge_attr="weight",
+            create_using=nx.DiGraph(),
+        )
+    assert nx.is_directed(nxG) is True
+    assert nx.is_weighted(nxG) is True
+
+    cuG, isNx = ensure_cugraph_obj_for_nx(nxG)
+    assert cugraph.is_directed(cuG) is True
+    assert cugraph.is_weighted(cuG) is True
+
+    # Get the modularity score for partitioning versus random assignment
+    cu_score = cugraph_call(cuG, partitions)
+    rand_score = random_call(cuG, partitions)
+
+    # Assert that the partitioning has better modularity than the random
+    # assignment
+    assert cu_score > rand_score
 
+
+@pytest.mark.parametrize("graph_file", utils.DATASETS)
+@pytest.mark.parametrize("partitions", PARTITIONS)
+def test_modularity_clustering_multi_column(graph_file, partitions):
     # Read in the graph and get a cugraph object
     cu_M = utils.read_csv_file(graph_file, read_weights_in_sp=False)
     cu_M.rename(columns={'0': 'src_0', '1': 'dst_0'}, inplace=True)
@@ -113,8 +141,6 @@ def test_modularity_clustering_multi_column(graph_file, partitions):
 
 
 def test_digraph_rejected():
-    gc.collect()
-
     df = cudf.DataFrame()
     df["src"] = cudf.Series(range(10))
     df["dst"] = cudf.Series(range(10))

python/cugraph/cugraph/tests/test_nx_convert.py

@@ -11,8 +11,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import gc
-
+import pandas as pd
 import pytest
 import cudf
 
@@ -39,13 +38,33 @@ def _compare_graphs(nxG, cuG, has_wt=True):
     cu_df = cuG.view_edge_list().to_pandas()
     if has_wt is True:
         cu_df = cu_df.drop(columns=["weights"])
+
+    out_of_order = cu_df[cu_df['src'] > cu_df['dst']]
+    if len(out_of_order) > 0:
+        out_of_order = out_of_order.rename(
+            columns={"src": "dst", "dst": "src"})
+        right_order = cu_df[cu_df['src'] < cu_df['dst']]
+        cu_df = pd.concat([out_of_order, right_order])
+        del out_of_order
+        del right_order
     cu_df = cu_df.sort_values(by=["src", "dst"]).reset_index(drop=True)
 
     nx_df = nx.to_pandas_edgelist(nxG)
     if has_wt is True:
         nx_df = nx_df.drop(columns=["weight"])
     nx_df = nx_df.rename(columns={"source": "src", "target": "dst"})
     nx_df = nx_df.astype('int32')
+
+    out_of_order = nx_df[nx_df['src'] > nx_df['dst']]
+    if len(out_of_order) > 0:
+        out_of_order = out_of_order.rename(
+                                    columns={"src": "dst", "dst": "src"})
+        right_order = nx_df[nx_df['src'] < nx_df['dst']]
+
+        nx_df = pd.concat([out_of_order, right_order])
+        del out_of_order
+        del right_order
+
     nx_df = nx_df.sort_values(by=["src", "dst"]).reset_index(drop=True)
 
     assert cu_df.to_dict() == nx_df.to_dict()
@@ -54,9 +73,6 @@ def _compare_graphs(nxG, cuG, has_wt=True):
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
 def test_networkx_compatibility(graph_file):
     # test to make sure cuGraph and Nx build similar Graphs
-
-    gc.collect()
-
     # Read in the graph
     M = utils.read_csv_for_nx(graph_file, read_weights_in_sp=True)
 
@@ -81,22 +97,52 @@ def test_networkx_compatibility(graph_file):
 
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
-def test_nx_convert(graph_file):
-    gc.collect()
-
+def test_nx_convert_undirected(graph_file):
     # read data and create a Nx Graph
     nx_df = utils.read_csv_for_nx(graph_file)
+    nxG = nx.from_pandas_edgelist(nx_df, "0", "1", create_using=nx.Graph)
+    assert nx.is_directed(nxG) is False
+    assert nx.is_weighted(nxG) is False
+
+    cuG = cugraph.utilities.convert_from_nx(nxG)
+    assert cuG.is_directed() is False
+    assert cuG.is_weighted() is False
+
+    _compare_graphs(nxG, cuG, has_wt=False)
+
+
+@pytest.mark.parametrize("graph_file", utils.DATASETS)
+def test_nx_convert_directed(graph_file):
+    # read data and create a Nx DiGraph
+    nx_df = utils.read_csv_for_nx(graph_file)
     nxG = nx.from_pandas_edgelist(nx_df, "0", "1", create_using=nx.DiGraph)
+    assert nxG.is_directed() is True
 
     cuG = cugraph.utilities.convert_from_nx(nxG)
+    assert cuG.is_directed() is True
+    assert cuG.is_weighted() is False
 
     _compare_graphs(nxG, cuG, has_wt=False)
 
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
-def test_nx_convert_multicol(graph_file):
-    gc.collect()
+def test_nx_convert_weighted(graph_file):
+    # read data and create a Nx DiGraph
+    nx_df = utils.read_csv_for_nx(graph_file, read_weights_in_sp=True)
+    nxG = nx.from_pandas_edgelist(nx_df, "0", "1", "weight",
+                                  create_using=nx.DiGraph)
+    assert nx.is_directed(nxG) is True
+    assert nx.is_weighted(nxG) is True
 
+    cuG = cugraph.utilities.convert_from_nx(nxG)
+    assert cugraph.is_directed(cuG) is True
+    assert cugraph.is_weighted(cuG) is True
+
+    _compare_graphs(nxG, cuG, has_wt=True)
+
+
+@pytest.mark.parametrize("graph_file", utils.DATASETS)
+def test_nx_convert_multicol(graph_file):
     # read data and create a Nx Graph
     nx_df = utils.read_csv_for_nx(graph_file)
 

python/cugraph/cugraph/utilities/nx_factory.py

@@ -28,20 +28,23 @@
 nx = import_optional("networkx")
 
 
-def convert_from_nx(nxG, weight=None):
+def convert_from_nx(nxG, weight=None, do_renumber=True):
     """
     weight, if given, is the string/name of the edge attr in nxG to use for
     weights in the resulting cugraph obj.  If nxG has no edge attributes,
     weight is ignored even if specified.
     """
-    if type(nxG) == nx.classes.graph.Graph:
+    if isinstance(nxG, nx.classes.digraph.DiGraph):
+        G = cugraph.Graph(directed=True)
+    elif isinstance(nxG, nx.classes.graph.Graph):
         G = cugraph.Graph()
-    elif type(nxG) == nx.classes.digraph.DiGraph:
-        G = cugraph.DiGraph()
     else:
-        raise ValueError("nxG does not appear to be a NetworkX graph type")
+        raise ValueError(
+            "nxG does not appear to be a supported NetworkX graph type")
 
+    is_weighted = nx.is_weighted(nxG)
     pdf = nx.to_pandas_edgelist(nxG)
+
     # Convert vertex columns to strings if they are not integers
     # This allows support for any vertex input type
     if pdf["source"].dtype not in [np.int32, np.int64] or \
@@ -54,20 +57,26 @@ def convert_from_nx(nxG, weight=None):
     if num_col < 2:
         raise ValueError("NetworkX graph did not contain edges")
 
-    if weight is None:
-        num_col == 2
+    if num_col == 2:
         pdf = pdf[["source", "target"]]
 
-    if num_col >= 3 and weight is not None:
-        pdf = pdf[["source", "target", weight]]
-        num_col = 3
+    if num_col >= 3:
+        if is_weighted is False:
+            pdf = pdf[["source", "target"]]
+        elif weight is None:
+            pdf = pdf[["source", "target", "weight"]]
+            weight = "weight"
+        else:
+            pdf = pdf[["source", "target", weight]]
 
     gdf = from_pandas(pdf)
 
     if num_col == 2:
-        G.from_cudf_edgelist(gdf, "source", "target")
+        G.from_cudf_edgelist(gdf, source="source", destination="target",
+                             renumber=do_renumber)
     else:
-        G.from_cudf_edgelist(gdf, "source", "target", weight)
+        G.from_cudf_edgelist(gdf, source="source", destination="target",
+                             edge_attr=weight, renumber=do_renumber)
 
     del gdf
     del pdf