Merge branch 'branch-23.12' into cugraph_dgl

ci/test_wheel_cugraph.sh

@@ -9,6 +9,6 @@ RAPIDS_PY_WHEEL_NAME="pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-whe
 python -m pip install --no-deps ./local-pylibcugraph-dep/pylibcugraph*.whl
 
 # Always install latest dask for testing
-python -m pip install git+https://github.com/dask/dask.git@2023.9.2 git+https://github.com/dask/distributed.git@2023.9.2
+python -m pip install git+https://github.com/dask/dask.git@main git+https://github.com/dask/distributed.git@main
 
 ./ci/test_wheel.sh cugraph python/cugraph

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -20,7 +20,7 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask-core==2023.9.2
+- dask-core>=2023.9.2
 - dask-cuda==23.12.*
 - dask-cudf==23.12.*
 - dask>=2023.7.1

conda/environments/all_cuda-120_arch-x86_64.yaml

@@ -20,7 +20,7 @@ dependencies:
 - cupy>=12.0.0
 - cxx-compiler
 - cython>=3.0.0
-- dask-core==2023.9.2
+- dask-core>=2023.9.2
 - dask-cuda==23.12.*
 - dask-cudf==23.12.*
 - dask>=2023.7.1

conda/recipes/cugraph-pyg/meta.yaml

@@ -26,7 +26,7 @@ requirements:
     - python
     - scikit-build >=0.13.1
   run:
-    - distributed ==2023.9.2
+    - distributed >=2023.9.2
     - numba >=0.57
     - numpy >=1.21
     - python

conda/recipes/cugraph-service/meta.yaml

@@ -59,7 +59,7 @@ outputs:
         - cupy >=12.0.0
         - dask-cuda ={{ minor_version }}
         - dask-cudf ={{ minor_version }}
-        - distributed ==2023.9.2
+        - distributed >=2023.9.2
         - numba >=0.57
         - numpy >=1.21
         - python

conda/recipes/cugraph/meta.yaml

@@ -76,9 +76,9 @@ requirements:
     - cupy >=12.0.0
     - dask-cuda ={{ minor_version }}
     - dask-cudf ={{ minor_version }}
-    - dask ==2023.9.2
-    - dask-core ==2023.9.2
-    - distributed ==2023.9.2
+    - dask >=2023.9.2
+    - dask-core >=2023.9.2
+    - distributed >=2023.9.2
     - fsspec>=0.6.0
     - libcugraph ={{ version }}
     - pylibcugraph ={{ version }}

dependencies.yaml

@@ -385,7 +385,7 @@ dependencies:
       - output_types: conda
         packages:
           - aiohttp
-          - &dask-core_conda dask-core==2023.9.2
+          - &dask-core_conda dask-core>=2023.9.2
           - fsspec>=0.6.0
           - libcudf==23.12.*
           - requests

python/nx-cugraph/_nx_cugraph/__init__.py

@@ -30,11 +30,16 @@
     "functions": {
         # BEGIN: functions
         "betweenness_centrality",
+        "degree_centrality",
         "edge_betweenness_centrality",
+        "in_degree_centrality",
         "is_isolate",
         "isolates",
+        "k_truss",
         "louvain_communities",
         "number_of_isolates",
+        "number_of_selfloops",
+        "out_degree_centrality",
         # END: functions
     },
     "extra_docstrings": {

python/nx-cugraph/lint.yaml

@@ -45,12 +45,12 @@ repos:
       - id: pyupgrade
         args: [--py39-plus]
   - repo: https://github.com/psf/black
-    rev: 23.9.1
+    rev: 23.10.0
     hooks:
       - id: black
       # - id: black-jupyter
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.0.292
+    rev: v0.1.1
     hooks:
       - id: ruff
         args: [--fix-only, --show-fixes]
@@ -77,7 +77,7 @@ repos:
         additional_dependencies: [tomli]
         files: ^(nx_cugraph|docs)/
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.0.292
+    rev: v0.1.1
     hooks:
       - id: ruff
   - repo: https://github.com/pre-commit/pre-commit-hooks

python/nx-cugraph/nx_cugraph/algorithms/__init__.py

@@ -12,4 +12,5 @@
 # limitations under the License.
 from . import centrality, community
 from .centrality import *
+from .core import *
 from .isolate import *

python/nx-cugraph/nx_cugraph/algorithms/centrality/__init__.py

@@ -11,3 +11,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 from .betweenness import *
+from .degree_alg import *

python/nx-cugraph/nx_cugraph/algorithms/centrality/degree_alg.py

@@ -0,0 +1,48 @@
+# Copyright (c) 2023, NVIDIA CORPORATION.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from nx_cugraph.convert import _to_directed_graph, _to_graph
+from nx_cugraph.utils import networkx_algorithm, not_implemented_for
+
+__all__ = ["degree_centrality", "in_degree_centrality", "out_degree_centrality"]
+
+
+@networkx_algorithm
+def degree_centrality(G):
+    G = _to_graph(G)
+    if len(G) <= 1:
+        return dict.fromkeys(G, 1)
+    deg = G._degrees_array()
+    centrality = deg * (1 / (len(G) - 1))
+    return G._nodearray_to_dict(centrality)
+
+
+@not_implemented_for("undirected")
+@networkx_algorithm
+def in_degree_centrality(G):
+    G = _to_directed_graph(G)
+    if len(G) <= 1:
+        return dict.fromkeys(G, 1)
+    deg = G._in_degrees_array()
+    centrality = deg * (1 / (len(G) - 1))
+    return G._nodearray_to_dict(centrality)
+
+
+@not_implemented_for("undirected")
+@networkx_algorithm
+def out_degree_centrality(G):
+    G = _to_directed_graph(G)
+    if len(G) <= 1:
+        return dict.fromkeys(G, 1)
+    deg = G._out_degrees_array()
+    centrality = deg * (1 / (len(G) - 1))
+    return G._nodearray_to_dict(centrality)

python/nx-cugraph/nx_cugraph/algorithms/community/louvain.py

@@ -42,7 +42,7 @@ def louvain_communities(
     # NetworkX allows both directed and undirected, but cugraph only allows undirected.
     seed = _seed_to_int(seed)  # Unused, but ensure it's valid for future compatibility
     G = _to_undirected_graph(G, weight)
-    if G.row_indices.size == 0:
+    if G.src_indices.size == 0:
         # TODO: PLC doesn't handle empty graphs gracefully!
         return [{key} for key in G._nodeiter_to_iter(range(len(G)))]
     if max_level is None:

python/nx-cugraph/nx_cugraph/algorithms/core.py

@@ -0,0 +1,96 @@
+# Copyright (c) 2023, NVIDIA CORPORATION.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import cupy as cp
+import networkx as nx
+import numpy as np
+import pylibcugraph as plc
+
+import nx_cugraph as nxcg
+from nx_cugraph.utils import networkx_algorithm, not_implemented_for
+
+__all__ = ["k_truss"]
+
+
+@not_implemented_for("directed")
+@not_implemented_for("multigraph")
+@networkx_algorithm
+def k_truss(G, k):
+    if is_nx := isinstance(G, nx.Graph):
+        G = nxcg.from_networkx(G, preserve_all_attrs=True)
+    if nxcg.number_of_selfloops(G) > 0:
+        raise nx.NetworkXError(
+            "Input graph has self loops which is not permitted; "
+            "Consider using G.remove_edges_from(nx.selfloop_edges(G))."
+        )
+    # TODO: create renumbering helper function(s)
+    if k < 3:
+        # k-truss graph is comprised of nodes incident on k-2 triangles, so k<3 is a
+        # boundary condition. Here, all we need to do is drop nodes with zero degree.
+        # Technically, it would be okay to delete this branch of code, because
+        # plc.k_truss_subgraph behaves the same for 0 <= k < 3. We keep this branch b/c
+        # it's faster and has an "early return" if there are no nodes with zero degree.
+        degrees = G._degrees_array()
+        # Renumber step 0: node indices
+        node_indices = degrees.nonzero()[0]
+        if degrees.size == node_indices.size:
+            # No change
+            return G if is_nx else G.copy()
+        src_indices = G.src_indices
+        dst_indices = G.dst_indices
+        # Renumber step 1: edge values (no changes needed)
+        edge_values = {key: val.copy() for key, val in G.edge_values.items()}
+        edge_masks = {key: val.copy() for key, val in G.edge_masks.items()}
+    else:
+        # int dtype for edge_indices would be preferred
+        edge_indices = cp.arange(G.src_indices.size, dtype=np.float64)
+        src_indices, dst_indices, edge_indices, _ = plc.k_truss_subgraph(
+            resource_handle=plc.ResourceHandle(),
+            graph=G._get_plc_graph(edge_array=edge_indices),
+            k=k,
+            do_expensive_check=False,
+        )
+        # Renumber step 0: node indices
+        node_indices = cp.unique(cp.concatenate([src_indices, dst_indices]))
+        # Renumber step 1: edge values
+        edge_indices = edge_indices.astype(np.int64)
+        edge_values = {key: val[edge_indices] for key, val in G.edge_values.items()}
+        edge_masks = {key: val[edge_indices] for key, val in G.edge_masks.items()}
+    # Renumber step 2: edge indices
+    mapper = cp.zeros(len(G), src_indices.dtype)
+    mapper[node_indices] = cp.arange(node_indices.size, dtype=mapper.dtype)
+    src_indices = mapper[src_indices]
+    dst_indices = mapper[dst_indices]
+    # Renumber step 3: node values
+    node_values = {key: val[node_indices] for key, val in G.node_values.items()}
+    node_masks = {key: val[node_indices] for key, val in G.node_masks.items()}
+    # Renumber step 4: key_to_id
+    if (id_to_key := G.id_to_key) is not None:
+        key_to_id = {
+            id_to_key[old_index]: new_index
+            for new_index, old_index in enumerate(node_indices.tolist())
+        }
+    else:
+        key_to_id = None
+    # Same as calling `G.from_coo`, but use __class__ to indicate it's a classmethod.
+    new_graph = G.__class__.from_coo(
+        node_indices.size,
+        src_indices,
+        dst_indices,
+        edge_values,
+        edge_masks,
+        node_values,
+        node_masks,
+        key_to_id=key_to_id,
+    )
+    new_graph.graph.update(G.graph)
+    return new_graph

python/nx-cugraph/nx_cugraph/algorithms/isolate.py

@@ -30,18 +30,18 @@ def is_isolate(G, n):
     G = _to_graph(G)
     index = n if G.key_to_id is None else G.key_to_id[n]
     return not (
-        (G.row_indices == index).any().tolist()
+        (G.src_indices == index).any().tolist()
         or G.is_directed()
-        and (G.col_indices == index).any().tolist()
+        and (G.dst_indices == index).any().tolist()
     )
 
 
 def _mark_isolates(G) -> cp.ndarray[bool]:
     """Return a boolean mask array indicating indices of isolated nodes."""
     mark_isolates = cp.ones(len(G), bool)
-    mark_isolates[G.row_indices] = False
+    mark_isolates[G.src_indices] = False
     if G.is_directed():
-        mark_isolates[G.col_indices] = False
+        mark_isolates[G.dst_indices] = False
     return mark_isolates
 
 

python/nx-cugraph/nx_cugraph/classes/__init__.py

@@ -11,7 +11,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 from .graph import Graph
+from .digraph import DiGraph
 from .multigraph import MultiGraph
+from .multidigraph import MultiDiGraph
 
-from .digraph import DiGraph  # isort:skip
-from .multidigraph import MultiDiGraph  # isort:skip
+from .function import *

python/nx-cugraph/nx_cugraph/classes/digraph.py

@@ -14,6 +14,7 @@
 
 from typing import TYPE_CHECKING
 
+import cupy as cp
 import networkx as nx
 
 import nx_cugraph as nxcg
@@ -48,7 +49,7 @@ def number_of_edges(
     ) -> int:
         if u is not None or v is not None:
             raise NotImplementedError
-        return self.row_indices.size
+        return self.src_indices.size
 
     ##########################
     # NetworkX graph methods #
@@ -59,3 +60,13 @@ def reverse(self, copy: bool = True) -> DiGraph:
         return self._copy(not copy, self.__class__, reverse=True)
 
     # Many more methods to implement...
+
+    ###################
+    # Private methods #
+    ###################
+
+    def _in_degrees_array(self):
+        return cp.bincount(self.dst_indices, minlength=self._N)
+
+    def _out_degrees_array(self):
+        return cp.bincount(self.src_indices, minlength=self._N)

python/nx-cugraph/nx_cugraph/classes/function.py

@@ -0,0 +1,23 @@
+# Copyright (c) 2023, NVIDIA CORPORATION.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from nx_cugraph.convert import _to_graph
+from nx_cugraph.utils import networkx_algorithm
+
+__all__ = ["number_of_selfloops"]
+
+
+@networkx_algorithm
+def number_of_selfloops(G):
+    G = _to_graph(G)
+    is_selfloop = G.src_indices == G.dst_indices
+    return is_selfloop.sum().tolist()