Update some clustering algos to only support undirected graphs (rapid…

…sai#2267) Our current implementation of `Louvain`, `Leiden`, `KTruss` and `Triangle Count` only support `undirected graphs`. To enforce that, we are currently raising an `exception` if a graph is not of `type Graph` but this is ineffective if a user create a `cugraph.Graph` with the parameter` directed=True` This PR - Raises an exception if `directed=True` when creating a `cugraph.Graph` - Update the docstrings for those algos - Add tests closes rapidsai#2263 Authors: - Joseph Nke (https://github.com/jnke2016) Approvers: - Brad Rees (https://github.com/BradReesWork) - Rick Ratzel (https://github.com/rlratzel) URL: rapidsai#2267
oorliu · May 17, 2022 · da3911c · da3911c
1 parent a908ccd
commit da3911c
Show file tree

Hide file tree

Showing 10 changed files with 126 additions and 53 deletions.
diff --git a/python/cugraph/cugraph/community/ktruss_subgraph.py b/python/cugraph/cugraph/community/ktruss_subgraph.py
@@ -134,6 +134,7 @@ def ktruss_subgraph(G, k, use_weights=True):
         cuGraph graph descriptor with connectivity information. k-Trusses are
         defined for only undirected graphs as they are defined for
         undirected triangle in a graph.
+        The current implementation only supports undirected graphs.
 
     k : int
         The desired k to be used for extracting the k-truss subgraph.
@@ -159,8 +160,8 @@ def ktruss_subgraph(G, k, use_weights=True):
     _ensure_compatible_cuda_version()
 
     KTrussSubgraph = Graph()
-    if type(G) is not Graph:
-        raise Exception("input graph must be undirected")
+    if G.is_directed():
+        raise ValueError("input graph must be undirected")
 
     subgraph_df = ktruss_subgraph_wrapper.ktruss_subgraph(G, k, use_weights)
     if G.renumbered:

diff --git a/python/cugraph/cugraph/community/leiden.py b/python/cugraph/cugraph/community/leiden.py
@@ -12,7 +12,6 @@
 # limitations under the License.
 
 from cugraph.community import leiden_wrapper
-from cugraph.structure.graph_classes import Graph
 from cugraph.utilities import (ensure_cugraph_obj_for_nx,
                                df_score_to_dictionary,
                                )
@@ -34,6 +33,8 @@ def leiden(G, max_iter=100, resolution=1.):
     G : cugraph.Graph
         cuGraph graph descriptor of type Graph
 
+        The current implementation only supports undirected graphs.
+
         The adjacency list will be computed if not already present.
 
     max_iter : integer, optional (default=100)
@@ -76,8 +77,8 @@ def leiden(G, max_iter=100, resolution=1.):
     """
     G, isNx = ensure_cugraph_obj_for_nx(G)
 
-    if type(G) is not Graph:
-        raise Exception(f"input graph must be undirected was {type(G)}")
+    if G.is_directed():
+        raise ValueError("input graph must be undirected")
 
     parts, modularity_score = leiden_wrapper.leiden(
         G, max_iter, resolution

diff --git a/python/cugraph/cugraph/community/louvain.py b/python/cugraph/cugraph/community/louvain.py
@@ -12,7 +12,6 @@
 # limitations under the License.
 
 from cugraph.community import louvain_wrapper
-from cugraph.structure.graph_classes import Graph
 from cugraph.utilities import (ensure_cugraph_obj_for_nx,
                                df_score_to_dictionary,
                                )
@@ -35,6 +34,7 @@ def louvain(G, max_iter=100, resolution=1.):
         The graph descriptor should contain the connectivity information
         and weights. The adjacency list will be computed if not already
         present.
+        The current implementation only supports undirected graphs.
 
     max_iter : integer, optional (default=100)
         This controls the maximum number of levels/iterations of the Louvain
@@ -77,8 +77,8 @@ def louvain(G, max_iter=100, resolution=1.):
 
     G, isNx = ensure_cugraph_obj_for_nx(G)
 
-    if type(G) is not Graph:
-        raise Exception("input graph must be undirected")
+    if G.is_directed():
+        raise ValueError("input graph must be undirected")
 
     parts, modularity_score = louvain_wrapper.louvain(
         G, max_iter, resolution

diff --git a/python/cugraph/cugraph/community/triangle_count.py b/python/cugraph/cugraph/community/triangle_count.py
@@ -12,7 +12,6 @@
 # limitations under the License.
 
 from cugraph.community import triangle_count_wrapper
-from cugraph.structure.graph_classes import Graph
 from cugraph.utilities import ensure_cugraph_obj_for_nx
 
 
@@ -28,7 +27,8 @@ def triangles(G):
     ----------
     G : cugraph.graph or networkx.Graph
         cuGraph graph descriptor, should contain the connectivity information,
-        (edge weights are not used in this algorithm)
+        (edge weights are not used in this algorithm).
+        The current implementation only supports undirected graphs.
 
     Returns
     -------
@@ -50,8 +50,8 @@ def triangles(G):
 
     G, _ = ensure_cugraph_obj_for_nx(G)
 
-    if type(G) is not Graph:
-        raise Exception("input graph must be undirected")
+    if G.is_directed():
+        raise ValueError("input graph must be undirected")
 
     result = triangle_count_wrapper.triangles(G)
 

diff --git a/python/cugraph/cugraph/dask/community/louvain.py b/python/cugraph/cugraph/dask/community/louvain.py
@@ -69,6 +69,8 @@ def louvain(input_graph, max_iter=100, resolution=1.0):
         and weights. The adjacency list will be computed if not already
         present.
 
+        The current implementation only supports undirected graphs.
+
     max_iter : integer, optional (default=100)
         This controls the maximum number of levels/iterations of the Louvain
         algorithm. When specified the algorithm will terminate after no more
@@ -114,11 +116,8 @@ def louvain(input_graph, max_iter=100, resolution=1.0):
     >>> parts, modularity_score = dcg.louvain(dg)
 
     """
-    # FIXME: dask methods to populate graphs from edgelists are only present on
-    # DiGraph classes. Disable the Graph check for now and assume inputs are
-    # symmetric DiGraphs.
-    # if type(graph) is not Graph:
-    #     raise Exception("input graph must be undirected")
+    if input_graph.is_directed():
+        raise ValueError("input graph must be undirected")
     client = default_client()
     # Calling renumbering results in data that is sorted by degree
     input_graph.compute_renumber_edge_list(transposed=False)

diff --git a/python/cugraph/cugraph/tests/mg/test_mg_louvain.py b/python/cugraph/cugraph/tests/mg/test_mg_louvain.py
@@ -37,19 +37,26 @@ def setFixtureParamNames(*args, **kwargs):
         pass
 
 
+# =============================================================================
+# Parameters
+# =============================================================================
+DATASETS_ASYMMETRIC = [
+    utils.RAPIDS_DATASET_ROOT_DIR_PATH/"karate-asymmetric.csv"]
+
+
 ###############################################################################
 # Fixtures
 # @pytest.mark.skipif(
 #    is_single_gpu(), reason="skipping MG testing on Single GPU system"
 # )
 @pytest.fixture(scope="module",
-                params=utils.DATASETS_UNDIRECTED,
+                params=DATASETS_ASYMMETRIC,
                 ids=[f"dataset={d.as_posix()}"
-                     for d in utils.DATASETS_UNDIRECTED])
+                     for d in DATASETS_ASYMMETRIC])
 def daskGraphFromDataset(request, dask_client):
     """
     Returns a new dask dataframe created from the dataset file param.
-    This creates un undirected Graph.
+    This creates a directed Graph.
     """
     # Since parameterized fixtures do not assign param names to param values,
     # manually call the helper to do so.
@@ -77,7 +84,7 @@ def daskGraphFromDataset(request, dask_client):
 def uddaskGraphFromDataset(request, dask_client):
     """
     Returns a new dask dataframe created from the dataset file param.
-    This creates un undirected Graph.
+    This creates an undirected Graph.
     """
     # Since parameterized fixtures do not assign param names to param
     # values, manually call the helper to do so.
@@ -103,26 +110,19 @@ def uddaskGraphFromDataset(request, dask_client):
 # @pytest.mark.skipif(
 #    is_single_gpu(), reason="skipping MG testing on Single GPU system"
 # )
-def test_mg_louvain_with_edgevals(daskGraphFromDataset):
-    # FIXME: daskGraphFromDataset returns a Directed graph, which Louvain is
-    # currently accepting. In the future, an MNMG symmeterize will need to
-    # be called to create a Graph for Louvain.
-    parts, mod = dcg.louvain(daskGraphFromDataset)
-
-    # FIXME: either call Nx with the same dataset and compare results, or
-    # hardcode golden results to compare to.
-    print()
-    print(parts.compute())
-    print(mod)
-    print()
+def test_mg_louvain_with_edgevals_directed_graph(daskGraphFromDataset):
+    # Directed graphs are not supported by Louvain and a ValueError should be
+    # raised
+    with pytest.raises(ValueError):
+        parts, mod = dcg.louvain(daskGraphFromDataset)
 
 
 ###############################################################################
 # Tests
 # @pytest.mark.skipif(
 #    is_single_gpu(), reason="skipping MG testing on Single GPU system"
 # )
-def test_mg_udlouvain_with_edgevals(uddaskGraphFromDataset):
+def test_mg_louvain_with_edgevals_undirected_graph(uddaskGraphFromDataset):
     parts, mod = dcg.louvain(uddaskGraphFromDataset)
 
     # FIXME: either call Nx with the same dataset and compare results, or

diff --git a/python/cugraph/cugraph/tests/test_k_truss_subgraph.py b/python/cugraph/cugraph/tests/test_k_truss_subgraph.py
@@ -35,6 +35,13 @@
 print("Networkx version : {} ".format(nx.__version__))
 
 
+# =============================================================================
+# Pytest Setup / Teardown - called for each test function
+# =============================================================================
+def setup_function():
+    gc.collect()
+
+
 # These ground truth files have been created by running the networkx ktruss
 # function on reference graphs. Currently networkx ktruss has an error such
 # that nx.k_truss(G,k-2) gives the expected result for running ktruss with
@@ -101,7 +108,6 @@ def test_unsupported_cuda_version():
                     f"{__unsupported_cuda_version} environment.")
 @pytest.mark.parametrize("graph_file, nx_ground_truth", utils.DATASETS_KTRUSS)
 def test_ktruss_subgraph_Graph(graph_file, nx_ground_truth):
-    gc.collect()
 
     k = 5
     cu_M = utils.read_csv_file(graph_file)
@@ -117,7 +123,6 @@ def test_ktruss_subgraph_Graph(graph_file, nx_ground_truth):
                     f"{__unsupported_cuda_version} environment.")
 @pytest.mark.parametrize("graph_file, nx_ground_truth", utils.DATASETS_KTRUSS)
 def test_ktruss_subgraph_Graph_nx(graph_file, nx_ground_truth):
-    gc.collect()
 
     k = 5
     M = utils.read_csv_for_nx(graph_file, read_weights_in_sp=True)
@@ -129,3 +134,22 @@ def test_ktruss_subgraph_Graph_nx(graph_file, nx_ground_truth):
     k_truss_nx = nx.k_truss(G, k)
 
     assert nx.is_isomorphic(k_subgraph, k_truss_nx)
+
+
+@pytest.mark.skipif((__cuda_version == __unsupported_cuda_version),
+                    reason="skipping on unsupported CUDA "
+                    f"{__unsupported_cuda_version} environment.")
+def test_ktruss_subgraph_directed_Graph():
+    input_data_path = (utils.RAPIDS_DATASET_ROOT_DIR_PATH /
+                       "karate-asymmetric.csv").as_posix()
+    k = 5
+    edgevals = True
+    cu_M = utils.read_csv_file(input_data_path)
+    G = cugraph.Graph(directed=True)
+    if edgevals:
+        G.from_cudf_edgelist(cu_M, source="0", destination="1", edge_attr="2")
+    else:
+        G.from_cudf_edgelist(cu_M, source="0", destination="1")
+
+    with pytest.raises(ValueError):
+        cugraph.k_truss(G, k)
diff --git a/python/cugraph/cugraph/tests/test_leiden.py b/python/cugraph/cugraph/tests/test_leiden.py
@@ -31,7 +31,14 @@
     warnings.filterwarnings("ignore", category=DeprecationWarning)
 
 
-def cugraph_leiden(G, edgevals=False):
+# =============================================================================
+# Pytest Setup / Teardown - called for each test function
+# =============================================================================
+def setup_function():
+    gc.collect()
+
+
+def cugraph_leiden(G):
 
     # cugraph Louvain Call
     t1 = time.time()
@@ -42,7 +49,7 @@ def cugraph_leiden(G, edgevals=False):
     return parts, mod
 
 
-def cugraph_louvain(G, edgevals=False):
+def cugraph_louvain(G):
 
     # cugraph Louvain Call
     t1 = time.time()
@@ -55,7 +62,6 @@ def cugraph_louvain(G, edgevals=False):
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
 def test_leiden(graph_file):
-    gc.collect()
     edgevals = True
 
     cu_M = utils.read_csv_file(graph_file)
@@ -66,16 +72,15 @@ def test_leiden(graph_file):
     else:
         G.from_cudf_edgelist(cu_M, source="0", destination="1")
 
-    leiden_parts, leiden_mod = cugraph_leiden(G, edgevals=True)
-    louvain_parts, louvain_mod = cugraph_louvain(G, edgevals=True)
+    leiden_parts, leiden_mod = cugraph_leiden(G)
+    louvain_parts, louvain_mod = cugraph_louvain(G)
 
     # Calculating modularity scores for comparison
     assert leiden_mod >= (0.99 * louvain_mod)
 
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS)
 def test_leiden_nx(graph_file):
-    gc.collect()
     edgevals = True
 
     NM = utils.read_csv_for_nx(graph_file)
@@ -89,8 +94,24 @@ def test_leiden_nx(graph_file):
             NM, create_using=nx.Graph(), source="0", target="1", edge_attr="2"
         )
 
-    leiden_parts, leiden_mod = cugraph_leiden(G, edgevals=True)
-    louvain_parts, louvain_mod = cugraph_louvain(G, edgevals=True)
+    leiden_parts, leiden_mod = cugraph_leiden(G)
+    louvain_parts, louvain_mod = cugraph_louvain(G)
 
     # Calculating modularity scores for comparison
     assert leiden_mod >= (0.99 * louvain_mod)
+
+
+def test_leiden_directed_graph():
+    input_data_path = (utils.RAPIDS_DATASET_ROOT_DIR_PATH /
+                       "karate-asymmetric.csv").as_posix()
+
+    edgevals = True
+    cu_M = utils.read_csv_file(input_data_path)
+    G = cugraph.Graph(directed=True)
+    if edgevals:
+        G.from_cudf_edgelist(cu_M, source="0", destination="1", edge_attr="2")
+    else:
+        G.from_cudf_edgelist(cu_M, source="0", destination="1")
+
+    with pytest.raises(ValueError):
+        parts, mod = cugraph_leiden(G)
diff --git a/python/cugraph/cugraph/tests/test_louvain.py b/python/cugraph/cugraph/tests/test_louvain.py
@@ -41,9 +41,16 @@
 print("Networkx version : {} ".format(nx.__version__))
 
 
-def cugraph_call(cu_M, edgevals=False):
+# =============================================================================
+# Pytest Setup / Teardown - called for each test function
+# =============================================================================
+def setup_function():
+    gc.collect()
+
+
+def cugraph_call(cu_M, edgevals=False, directed=False):
 
-    G = cugraph.Graph()
+    G = cugraph.Graph(directed=directed)
     if edgevals:
         G.from_cudf_edgelist(cu_M, source="0", destination="1", edge_attr="2")
     else:
@@ -75,7 +82,6 @@ def networkx_call(M):
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS_UNDIRECTED)
 def test_louvain_with_edgevals(graph_file):
-    gc.collect()
 
     M = utils.read_csv_for_nx(graph_file)
     cu_M = utils.read_csv_file(graph_file)
@@ -103,7 +109,6 @@ def test_louvain_with_edgevals(graph_file):
 
 @pytest.mark.parametrize("graph_file", utils.DATASETS_UNDIRECTED)
 def test_louvain(graph_file):
-    gc.collect()
 
     M = utils.read_csv_for_nx(graph_file)
     cu_M = utils.read_csv_file(graph_file)
@@ -127,3 +132,13 @@ def test_louvain(graph_file):
     assert len(cu_parts) == len(nx_parts)
     assert cu_mod > (0.82 * nx_mod)
     assert abs(cu_mod - cu_mod_nx) < 0.0001
+
+
+def test_louvain_directed_graph():
+    input_data_path = (utils.RAPIDS_DATASET_ROOT_DIR_PATH /
+                       "karate-asymmetric.csv").as_posix()
+
+    cu_M = utils.read_csv_file(input_data_path)
+
+    with pytest.raises(ValueError):
+        cu_parts, cu_mod = cugraph_call(cu_M, directed=True)