Refactor cudf::contains

conda/recipes/libcudf/meta.yaml

@@ -247,7 +247,6 @@ outputs:
         - test -f $PREFIX/include/cudf/structs/structs_column_view.hpp
         - test -f $PREFIX/include/cudf/structs/struct_view.hpp
         - test -f $PREFIX/include/cudf/structs/detail/concatenate.hpp
-        - test -f $PREFIX/include/cudf/structs/detail/contains.hpp
         - test -f $PREFIX/include/cudf/table/table.hpp
         - test -f $PREFIX/include/cudf/table/table_view.hpp
         - test -f $PREFIX/include/cudf/tdigest/tdigest_column_view.cuh

cpp/CMakeLists.txt

@@ -446,6 +446,7 @@ add_library(
   src/scalar/scalar.cpp
   src/scalar/scalar_factories.cpp
   src/search/contains.cu
+  src/search/contains_nested.cu
   src/search/search_ordered.cu
   src/sort/is_sorted.cu
   src/sort/rank.cu
@@ -514,7 +515,6 @@ add_library(
   src/strings/utilities.cu
   src/strings/wrap.cu
   src/structs/copying/concatenate.cu
-  src/structs/search/contains.cu
   src/structs/structs_column_factories.cu
   src/structs/structs_column_view.cpp
   src/structs/utilities.cpp

cpp/include/cudf/detail/search.hpp

@@ -23,35 +23,30 @@
 
 #include <rmm/cuda_stream_view.hpp>
 
-#include <vector>
-
-namespace cudf {
-namespace detail {
+namespace cudf::detail {
 /**
  * @copydoc cudf::lower_bound
  *
  * @param stream CUDA stream used for device memory operations and kernel launches.
  */
-std::unique_ptr<column> lower_bound(
-  table_view const& haystack,
-  table_view const& needles,
-  std::vector<order> const& column_order,
-  std::vector<null_order> const& null_precedence,
-  rmm::cuda_stream_view stream,
-  rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
+std::unique_ptr<column> lower_bound(table_view const& haystack,
+                                    table_view const& needles,
+                                    std::vector<order> const& column_order,
+                                    std::vector<null_order> const& null_precedence,
+                                    rmm::cuda_stream_view stream,
+                                    rmm::mr::device_memory_resource* mr);
 
 /**
  * @copydoc cudf::upper_bound
  *
  * @param stream CUDA stream used for device memory operations and kernel launches.
  */
-std::unique_ptr<column> upper_bound(
-  table_view const& haystack,
-  table_view const& needles,
-  std::vector<order> const& column_order,
-  std::vector<null_order> const& null_precedence,
-  rmm::cuda_stream_view stream,
-  rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
+std::unique_ptr<column> upper_bound(table_view const& haystack,
+                                    table_view const& needles,
+                                    std::vector<order> const& column_order,
+                                    std::vector<null_order> const& null_precedence,
+                                    rmm::cuda_stream_view stream,
+                                    rmm::mr::device_memory_resource* mr);
 
 /**
  * @copydoc cudf::contains(column_view const&, scalar const&, rmm::mr::device_memory_resource*)
@@ -61,15 +56,28 @@ std::unique_ptr<column> upper_bound(
 bool contains(column_view const& haystack, scalar const& needle, rmm::cuda_stream_view stream);
 
 /**
- * @copydoc cudf::contains(column_view const&, column_view const&, rmm::mr::device_memory_resource*)
- *
+
  * @param stream CUDA stream used for device memory operations and kernel launches.
  */
-std::unique_ptr<column> contains(
-  column_view const& haystack,
-  column_view const& needles,
-  rmm::cuda_stream_view stream,
-  rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource());
+std::unique_ptr<column> contains(column_view const& haystack,
+                                 column_view const& needles,
+                                 rmm::cuda_stream_view stream,
+                                 rmm::mr::device_memory_resource* mr);
+
+/**
+ * @brief Check if the (unique) row of the `needle` column is contained in the `haystack` column.
+ *
+ * If the input `needle` column has more than one row, only the first row will be considered.
+ *
+ * This function is designed for nested types only. It can also work with non-nested types
+ * but with lower performance due to the complexity of the implementation.
+ *
+ * @param haystack The column containing search space.
+ * @param needle A scalar value to check for existence in the search space.
+ * @return true if the given `needle` value exists in the `haystack` column.
+ */
+bool contains_nested_element(column_view const& haystack,
+                             column_view const& needle,
+                             rmm::cuda_stream_view stream);
 
-}  // namespace detail
-}  // namespace cudf
+}  // namespace cudf::detail

cpp/src/hash/unordered_multiset.cuh

@@ -131,7 +131,7 @@ class unordered_multiset {
     return unordered_multiset(d_col.size(), std::move(hash_bins_start), std::move(hash_data));
   }
 
-  unordered_multiset_device_view<Element, Hasher, Equality> to_device()
+  unordered_multiset_device_view<Element, Hasher, Equality> to_device() const
   {
     return unordered_multiset_device_view<Element, Hasher, Equality>(
       size, hash_bins.data(), hash_data.data());