make_empty_tdigests_column

cpp/include/cudf/detail/tdigest/tdigest.hpp

@@ -143,19 +143,20 @@ std::unique_ptr<column> make_tdigest_column(size_type num_rows,
                                             rmm::device_async_resource_ref mr);
 
 /**
- * @brief Create a tdigest column of empty clusters.
+ * @brief Create a tdigest column of empty tdigests.
  *
- * The column created contains the specified number of rows of empty clusters.
+ * The column created contains the specified number of rows of empty tdigests.
  *
+ * @param num_rows The number of rows in the output column.
  * @param stream CUDA stream used for device memory operations and kernel launches.
  * @param mr Device memory resource used to allocate the returned column's device memory.
  *
  * @returns A tdigest column of empty clusters.
  */
 CUDF_EXPORT
-std::unique_ptr<column> make_tdigest_column_of_empty_clusters(size_type num_rows,
-                                                              rmm::cuda_stream_view stream,
-                                                              rmm::device_async_resource_ref mr);
+std::unique_ptr<column> make_empty_tdigests_column(size_type num_rows,
+                                                   rmm::cuda_stream_view stream,
+                                                   rmm::device_async_resource_ref mr);
 
 /**
  * @brief Create a scalar of an empty tdigest cluster.

cpp/include/cudf_test/tdigest_utilities.cuh

@@ -270,7 +270,7 @@ void tdigest_simple_all_nulls_aggregation(Func op)
     static_cast<column_view>(values).type(), tdigest_gen{}, op, values, delta);
 
   // NOTE: an empty tdigest column still has 1 row.
-  auto expected = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto expected = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
 
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(*result, *expected);
@@ -562,11 +562,11 @@ template <typename MergeFunc>
 void tdigest_merge_empty(MergeFunc merge_op)
 {
   // 3 empty tdigests all in the same group
-  auto a = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto a = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
-  auto b = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto b = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
-  auto c = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto c = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
   std::vector<column_view> cols;
   cols.push_back(*a);
@@ -577,7 +577,7 @@ void tdigest_merge_empty(MergeFunc merge_op)
   auto const delta = 1000;
   auto result      = merge_op(*values, delta);
 
-  auto expected = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto expected = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
 
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(*expected, *result);

cpp/src/quantiles/tdigest/tdigest.cu

@@ -292,9 +292,9 @@ std::unique_ptr<column> make_tdigest_column(size_type num_rows,
   return make_structs_column(num_rows, std::move(children), 0, {}, stream, mr);
 }
 
-std::unique_ptr<column> make_tdigest_column_of_empty_clusters(size_type num_rows,
-                                                              rmm::cuda_stream_view stream,
-                                                              rmm::device_async_resource_ref mr)
+std::unique_ptr<column> make_empty_tdigests_column(size_type num_rows,
+                                                   rmm::cuda_stream_view stream,
+                                                   rmm::device_async_resource_ref mr)
 {
   auto offsets = cudf::make_fixed_width_column(
     data_type(type_id::INT32), num_rows + 1, mask_state::UNALLOCATED, stream, mr);
@@ -339,7 +339,7 @@ std::unique_ptr<column> make_tdigest_column_of_empty_clusters(size_type num_rows
 std::unique_ptr<scalar> make_empty_tdigest_scalar(rmm::cuda_stream_view stream,
                                                   rmm::device_async_resource_ref mr)
 {
-  auto contents = make_tdigest_column_of_empty_clusters(1, stream, mr)->release();
+  auto contents = make_empty_tdigests_column(1, stream, mr)->release();
   return std::make_unique<struct_scalar>(
     std::move(*std::make_unique<table>(std::move(contents.children))), true, stream, mr);
 }

cpp/src/quantiles/tdigest/tdigest_aggregation.cu

@@ -759,7 +759,7 @@ std::unique_ptr<column> compute_tdigests(int delta,
   // }
   //
   if (total_clusters == 0) {
-    return cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(1, stream, mr);
+    return cudf::tdigest::detail::make_empty_tdigests_column(1, stream, mr);
   }
 
   // each input group represents an individual tdigest.  within each tdigest, we want the keys
@@ -1339,9 +1339,7 @@ std::unique_ptr<column> group_tdigest(column_view const& col,
                                       rmm::cuda_stream_view stream,
                                       rmm::device_async_resource_ref mr)
 {
-  if (col.size() == 0) {
-    return cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(1, stream, mr);
-  }
+  if (col.size() == 0) { return cudf::tdigest::detail::make_empty_tdigests_column(1, stream, mr); }
 
   auto const delta = max_centroids;
   return cudf::type_dispatcher(col.type(),
@@ -1367,15 +1365,15 @@ std::unique_ptr<column> group_merge_tdigest(column_view const& input,
   tdigest_column_view tdv(input);
 
   if (num_groups == 0 || input.size() == 0) {
-    return cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(1, stream, mr);
+    return cudf::tdigest::detail::make_empty_tdigests_column(1, stream, mr);
   }
 
   if (tdv.means().size() == 0) {
     // `group_merge_tdigest` takes the output of `typed_group_tdigest` as its input, which wipes
     // out the means and weights for empty clusters. Thus, no mean here indicates that all clusters
     // are empty in the input. Let's skip all complex computation in the below, but just return
     // an empty tdigest per group.
-    return cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(num_groups, stream, mr);
+    return cudf::tdigest::detail::make_empty_tdigests_column(num_groups, stream, mr);
   }
 
   // bring group offsets back to the host

cpp/tests/groupby/tdigest_tests.cu

@@ -469,15 +469,15 @@ TEST_F(TDigestMergeTest, EmptyGroups)
   cudf::test::fixed_width_column_wrapper<int> keys{0, 0, 0, 0, 0, 0, 0};
   int const delta = 1000;
 
-  auto a = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto a = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
   auto b = cudf::type_dispatcher(
     static_cast<cudf::column_view>(values_b).type(), tdigest_gen_grouped{}, keys, values_b, delta);
-  auto c = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto c = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
   auto d = cudf::type_dispatcher(
     static_cast<cudf::column_view>(values_d).type(), tdigest_gen_grouped{}, keys, values_d, delta);
-  auto e = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto e = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
 
   std::vector<cudf::column_view> cols;
@@ -561,10 +561,10 @@ TEST_F(TDigestMergeTest, AllValuesAreNull)
 
   auto const expected_computed_keys = cudf::test::fixed_width_column_wrapper<int32_t>{{0, 1, 2}};
   cudf::column_view const expected_computed_keys_view{expected_computed_keys};
-  auto const expected_computed_vals = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
-    expected_computed_keys_view.size(),
-    cudf::get_default_stream(),
-    rmm::mr::get_current_device_resource());
+  auto const expected_computed_vals =
+    cudf::tdigest::detail::make_empty_tdigests_column(expected_computed_keys_view.size(),
+                                                      cudf::get_default_stream(),
+                                                      rmm::mr::get_current_device_resource());
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(expected_computed_keys_view, compute_result->get_column(0).view());
   // The computed values are nullable even though the input values are not.
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(expected_computed_vals->view(),
@@ -578,10 +578,10 @@ TEST_F(TDigestMergeTest, AllValuesAreNull)
 
   auto const expected_merged_keys = cudf::test::fixed_width_column_wrapper<int32_t>{{0, 1, 2}};
   cudf::column_view const expected_merged_keys_view{expected_merged_keys};
-  auto const expected_merged_vals = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
-    expected_merged_keys_view.size(),
-    cudf::get_default_stream(),
-    rmm::mr::get_current_device_resource());
+  auto const expected_merged_vals =
+    cudf::tdigest::detail::make_empty_tdigests_column(expected_merged_keys_view.size(),
+                                                      cudf::get_default_stream(),
+                                                      rmm::mr::get_current_device_resource());
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(expected_merged_keys_view, merge_result->get_column(0).view());
   CUDF_TEST_EXPECT_COLUMNS_EQUAL(expected_merged_vals->view(), merge_result->get_column(1).view());
 }

cpp/tests/quantiles/percentile_approx_test.cpp

@@ -371,7 +371,7 @@ struct PercentileApproxTest : public cudf::test::BaseFixture {};
 
 TEST_F(PercentileApproxTest, EmptyInput)
 {
-  auto empty_ = cudf::tdigest::detail::make_tdigest_column_of_empty_clusters(
+  auto empty_ = cudf::tdigest::detail::make_empty_tdigests_column(
     1, cudf::get_default_stream(), cudf::get_current_device_resource_ref());
   cudf::test::fixed_width_column_wrapper<double> percentiles{0.0, 0.25, 0.3};