Merge remote-tracking branch 'upstream/branch-25.08' into bc_scaling_to_match_nx

Author: rlratzel

CHANGELOG.md

@@ -1,3 +1,68 @@
+# cugraph 25.06.00 (5 Jun 2025)
+
+## 🚨 Breaking Changes
+
+- Convert non-type template parameter store_transposed in edge_src|dst_property_t to a runtime parameter ([#5074](https://github.com/rapidsai/cugraph/pull/5074)) [@seunghwak](https://github.com/seunghwak)
+- Remove Remaining cuGraph-DGL Code ([#5061](https://github.com/rapidsai/cugraph/pull/5061)) [@alexbarghi-nv](https://github.com/alexbarghi-nv)
+- Simplify edge_property_t ([#5053](https://github.com/rapidsai/cugraph/pull/5053)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Add support for random state to FA2 ([#5052](https://github.com/rapidsai/cugraph/pull/5052)) [@jnke2016](https://github.com/jnke2016)
+
+## 🐛 Bug Fixes
+
+- Adds `test_node2vec` to temporary non-xdist test list ([#5084](https://github.com/rapidsai/cugraph/pull/5084)) [@rlratzel](https://github.com/rlratzel)
+- Update method retrieving the edgelist ([#5082](https://github.com/rapidsai/cugraph/pull/5082)) [@jnke2016](https://github.com/jnke2016)
+- Pin cython to avoid version 3.1.0 because of bug (cython issue 6841) ([#5068](https://github.com/rapidsai/cugraph/pull/5068)) [@rlratzel](https://github.com/rlratzel)
+- Fix MSBFS Test Failure ([#5051](https://github.com/rapidsai/cugraph/pull/5051)) [@jnke2016](https://github.com/jnke2016)
+- Temporarily disable test so we can upgrade to CCCL 2.8.x ([#5047](https://github.com/rapidsai/cugraph/pull/5047)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Added  #include <cuda/std/cmath> for correct cuda::std::log reference ([#5039](https://github.com/rapidsai/cugraph/pull/5039)) [@SagaciousFish](https://github.com/SagaciousFish)
+- Fix ForceAtlas2 incorrect loop conditions ([#5030](https://github.com/rapidsai/cugraph/pull/5030)) [@MathisHammel](https://github.com/MathisHammel)
+- Fix sampling primitive build error. ([#5004](https://github.com/rapidsai/cugraph/pull/5004)) [@seunghwak](https://github.com/seunghwak)
+
+## 🛠️ Improvements
+
+- Test WCC with edge masking ([#5093](https://github.com/rapidsai/cugraph/pull/5093)) [@seunghwak](https://github.com/seunghwak)
+- use 'rapids-init-pip' in wheel CI, other CI changes ([#5089](https://github.com/rapidsai/cugraph/pull/5089)) [@jameslamb](https://github.com/jameslamb)
+- Removes unnecessary dependencies from cugraph-service ([#5087](https://github.com/rapidsai/cugraph/pull/5087)) [@rlratzel](https://github.com/rlratzel)
+- Utility functions to create an edge_src|dst_property view objects. ([#5078](https://github.com/rapidsai/cugraph/pull/5078)) [@seunghwak](https://github.com/seunghwak)
+- Test code for Graph 500 BFS ([#5077](https://github.com/rapidsai/cugraph/pull/5077)) [@seunghwak](https://github.com/seunghwak)
+- Remove pure=False from client.compute calls ([#5076](https://github.com/rapidsai/cugraph/pull/5076)) [@TomAugspurger](https://github.com/TomAugspurger)
+- Finish CUDA 12.9 migration and use branch-25.06 workflows ([#5075](https://github.com/rapidsai/cugraph/pull/5075)) [@bdice](https://github.com/bdice)
+- Convert non-type template parameter store_transposed in edge_src|dst_property_t to a runtime parameter ([#5074](https://github.com/rapidsai/cugraph/pull/5074)) [@seunghwak](https://github.com/seunghwak)
+- Update to clang 20 ([#5073](https://github.com/rapidsai/cugraph/pull/5073)) [@bdice](https://github.com/bdice)
+- Remove get_test_data before tests run. ([#5072](https://github.com/rapidsai/cugraph/pull/5072)) [@bdice](https://github.com/bdice)
+- Make thrust_tuple_cat device callable. ([#5070](https://github.com/rapidsai/cugraph/pull/5070)) [@seunghwak](https://github.com/seunghwak)
+- Quote head_rev in conda recipes ([#5069](https://github.com/rapidsai/cugraph/pull/5069)) [@bdice](https://github.com/bdice)
+- fix(rattler): add librmm to host to fix overlinking errors ([#5065](https://github.com/rapidsai/cugraph/pull/5065)) [@gforsyth](https://github.com/gforsyth)
+- CUDA 12.9 use updated compression flags ([#5064](https://github.com/rapidsai/cugraph/pull/5064)) [@robertmaynard](https://github.com/robertmaynard)
+- Remove Remaining cuGraph-DGL Code ([#5061](https://github.com/rapidsai/cugraph/pull/5061)) [@alexbarghi-nv](https://github.com/alexbarghi-nv)
+- Exclude librmm.so from auditwheel ([#5058](https://github.com/rapidsai/cugraph/pull/5058)) [@bdice](https://github.com/bdice)
+- Add support for Python 3.13 ([#5057](https://github.com/rapidsai/cugraph/pull/5057)) [@gforsyth](https://github.com/gforsyth)
+- Simplify edge_property_t ([#5053](https://github.com/rapidsai/cugraph/pull/5053)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Add support for random state to FA2 ([#5052](https://github.com/rapidsai/cugraph/pull/5052)) [@jnke2016](https://github.com/jnke2016)
+- [BUGFIX] Fixed quoting in wheel paths in cugraph and pylibcugraph tests ([#5049](https://github.com/rapidsai/cugraph/pull/5049)) [@VenkateshJaya](https://github.com/VenkateshJaya)
+- Fix failing test with new cccl ([#5046](https://github.com/rapidsai/cugraph/pull/5046)) [@ChuckHastings](https://github.com/ChuckHastings)
+- add C implementation of MST ([#5044](https://github.com/rapidsai/cugraph/pull/5044)) [@jnke2016](https://github.com/jnke2016)
+- Download build artifacts from GitHub for CI ([#5043](https://github.com/rapidsai/cugraph/pull/5043)) [@VenkateshJaya](https://github.com/VenkateshJaya)
+- Avoid unnecessary libcugraph wheel builds ([#5037](https://github.com/rapidsai/cugraph/pull/5037)) [@VenkateshJaya](https://github.com/VenkateshJaya)
+- Use public pylibcudf APIs in graph_primtypes ([#5036](https://github.com/rapidsai/cugraph/pull/5036)) [@mroeschke](https://github.com/mroeschke)
+- refactor(rattler): enable strict channel priority for builds ([#5035](https://github.com/rapidsai/cugraph/pull/5035)) [@gforsyth](https://github.com/gforsyth)
+- Reduce peak memory usage ([#5031](https://github.com/rapidsai/cugraph/pull/5031)) [@seunghwak](https://github.com/seunghwak)
+- Avoid private cuDF Python APIs ([#5028](https://github.com/rapidsai/cugraph/pull/5028)) [@mroeschke](https://github.com/mroeschke)
+- Use pytest-xdist for parallel testing ([#5027](https://github.com/rapidsai/cugraph/pull/5027)) [@bdice](https://github.com/bdice)
+- Vendor RAPIDS.cmake ([#5024](https://github.com/rapidsai/cugraph/pull/5024)) [@bdice](https://github.com/bdice)
+- Add missing thrust headers & Pass dtype objects to cuDF as_column ([#5023](https://github.com/rapidsai/cugraph/pull/5023)) [@alliepiper](https://github.com/alliepiper)
+- Add ARM conda environments ([#5021](https://github.com/rapidsai/cugraph/pull/5021)) [@bdice](https://github.com/bdice)
+- Add missing shuffle function declarations in shuffle_funcstions.hpp ([#5018](https://github.com/rapidsai/cugraph/pull/5018)) [@seunghwak](https://github.com/seunghwak)
+- Do not pretend to default initialize a device lambda ([#5015](https://github.com/rapidsai/cugraph/pull/5015)) [@miscco](https://github.com/miscco)
+- Update C++ test binary file output directory from build/tests to build/gtests ([#5014](https://github.com/rapidsai/cugraph/pull/5014)) [@seunghwak](https://github.com/seunghwak)
+- Replace `Thrust` iterator facilities with `libcu++` ones ([#5013](https://github.com/rapidsai/cugraph/pull/5013)) [@miscco](https://github.com/miscco)
+- Branch 25.06 merge 25.04 ([#5007](https://github.com/rapidsai/cugraph/pull/5007)) [@rlratzel](https://github.com/rlratzel)
+- Fixes Failing Cugraph Uniform Neighbor Sample Benchmark ([#5001](https://github.com/rapidsai/cugraph/pull/5001)) [@nv-rliu](https://github.com/nv-rliu)
+- Port all conda recipes to `rattler-build` ([#4999](https://github.com/rapidsai/cugraph/pull/4999)) [@gforsyth](https://github.com/gforsyth)
+- Moving wheel builds to specified location and uploading build artifacts to Github ([#4993](https://github.com/rapidsai/cugraph/pull/4993)) [@VenkateshJaya](https://github.com/VenkateshJaya)
+- Refactor Python Neighborhood Sample ([#4988](https://github.com/rapidsai/cugraph/pull/4988)) [@jnke2016](https://github.com/jnke2016)
+- Add shellcheck to pre-commit and fix warnings ([#4954](https://github.com/rapidsai/cugraph/pull/4954)) [@gforsyth](https://github.com/gforsyth)
+
 # cugraph 25.04.00 (9 Apr 2025)
 
 ## 🚨 Breaking Changes

ci/run_cugraph_pytests.sh

@@ -11,6 +11,5 @@ DASK_DISTRIBUTED__SCHEDULER__WORKER_TTL="${DASK_DISTRIBUTED__SCHEDULER__WORKER_T
 DASK_DISTRIBUTED__COMM__TIMEOUTS__CONNECT="${DASK_DISTRIBUTED__COMM__TIMEOUTS__CONNECT:-1000s}" \
 DASK_CUDA_WAIT_WORKERS_MIN_TIMEOUT="${DASK_CUDA_WAIT_WORKERS_MIN_TIMEOUT:-1000s}" \
 pytest --cache-clear --import-mode=append --benchmark-disable \
-  -k "not test_property_graph_mg" \
   "$@" \
   tests

ci/test_python.sh

@@ -82,31 +82,43 @@ rapids-logger "pytest cugraph (not mg, with xdist)"
   --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-coverage.xml" \
   --cov-report=term
 
-# Some tests fail with pytest-xdist enabled.
-# See https://github.com/rapidsai/cugraph/issues/5048
-rapids-logger "pytest cugraph (not mg, without xdist)"
+# The datasets tests may modify global states and interfere with tests running on other workers
+# these tests are typically only limited by network speed and run fairly quickly, so they can be
+# run separately
+rapids-logger "pytest cugraph (datasets APIs)"
 ./ci/run_cugraph_pytests.sh \
   --verbose \
   --junitxml="${RAPIDS_TESTS_DIR}/junit-cugraph.xml" \
-  --numprocesses=1 \
-  --dist=worksteal \
-  -m "not mg" \
-  -k "not test_dataset and (test_bulk_sampler or test_create_undirected_graph_from_asymmetric_adj_list or test_uniform_neighbor_sample or test_node2vec)" \
+  -k "test_dataset" \
   --cov-config=../../.coveragerc \
   --cov=cugraph \
   --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-coverage.xml" \
   --cov-report=term
 
-rapids-logger "pytest cugraph (mg)"
+# excludes known failures that will always fail when run in combination
+rapids-logger "pytest cugraph (mg, with xdist)"
 ./ci/run_cugraph_pytests.sh \
   --verbose \
   --junitxml="${RAPIDS_TESTS_DIR}/junit-cugraph.xml" \
+  --numprocesses=8 \
+  --dist=worksteal \
   -m "mg" \
+  -k "not test_dataset and not test_property_graph_mg and not test_dist_sampler_mg and not test_uniform_neighbor_sample_mg" \
   --cov-config=../../.coveragerc \
   --cov=cugraph \
   --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-coverage.xml" \
   --cov-report=term
 
+rapids-logger "pytest cugraph (mg dist_sampler and uns)"
+./ci/run_cugraph_pytests.sh \
+  --verbose \
+  --junitxml="${RAPIDS_TESTS_DIR}/junit-cugraph.xml" \
+  -m "mg" \
+  -k "test_dist_sampler_mg or test_uniform_neighbor_sample_mg" \
+  --cov-config=../../.coveragerc \
+  --cov=cugraph \
+  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cugraph-coverage.xml" \
+  --cov-report=term
 
 rapids-logger "pytest cugraph benchmarks (run as tests)"
 ./ci/run_cugraph_benchmark_pytests.sh --verbose

cpp/CMakeLists.txt

@@ -555,6 +555,7 @@ add_library(cugraph_c
         src/c_api/renumber_arbitrary_edgelist.cu
         src/c_api/legacy_fa2.cpp
         src/c_api/legacy_mst.cpp
+        src/c_api/extract_vertex_list.cpp
         )
 add_library(cugraph::cugraph_c ALIAS cugraph_c)
 

cpp/include/cugraph/algorithms.hpp

@@ -436,7 +436,6 @@ enum class cugraph_cc_t {
  * @ingroup components_cpp
  * @brief      Compute connected components.
  *
- * The weak version (for undirected graphs, only) was imported from cuML.
  * This implementation comes from [1] and solves component labeling problem in
  * parallel on CSR-indexes based upon the vertex degree and adjacency graph.
  *
@@ -1782,8 +1781,8 @@ enum class k_core_degree_type_t { IN = 0, OUT = 1, INOUT = 2 };
 .* @ingroup core_cpp
  * @brief   Compute core numbers of individual vertices from K-Core decomposition.
  *
- * The input graph should not have self-loops nor multi-edges. Currently, only undirected graphs are
- * supported.
+ * This algorithms does not support multi-graphs. Self-loops are excluded in computing core
+nuumbers.
  *
  * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
  * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
@@ -1814,7 +1813,9 @@ void core_number(raft::handle_t const& handle,
 .* @ingroup core_cpp
  * @brief   Extract K-Core of a graph
  *
- * @throws     cugraph::logic_error when an error occurs.
+ * This function internally calls core_number (if @p core_numbers.has_value() is false). core_number
+does not support multi-graphs. Self-loops are excluded in computing core nuumbers. Note that the
+extracted K-Core can still include self-loops.
  *
  * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
  * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
@@ -1851,6 +1852,9 @@ k_core(raft::handle_t const& handle,
  * Compute triangle counts for the entire set of vertices (if @p vertices is std::nullopt) or the
  * given vertices (@p vertices.has_value() is true).
  *
+ * This algorithms does not support multi-graphs. Self-loops are excluded in computing triangle
+ * counts.
+ *
  * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
  * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
  * @tparam multi_gpu Flag indicating whether template instantiation should target single-GPU (false)
@@ -1876,6 +1880,9 @@ void triangle_count(raft::handle_t const& handle,
  * @brief Compute edge triangle counts.
  *
  * Compute edge triangle counts for the entire set of edges.
+ *
+ * This algorithms does not support multi-graphs. Self-loops are excluded in computing edge triangle
+counts (they will have a triangle count of 0).
  *
  * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
  * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
@@ -1900,6 +1907,8 @@ edge_property_t<edge_t, edge_t> edge_triangle_count(
  *
  * Extract the K-Truss subgraph of a graph
  *
+ * This algorithms does not support multi-graphs. Self-loops are excluded in computing K-Truss.
+ *
  * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
  * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
  * @tparam multi_gpu Flag indicating whether template instantiation should target single-GPU (false)