[REVIEW] Fix cudf::strings:split logic for many columns

CHANGELOG.md

@@ -180,6 +180,7 @@
 - PR #4915 Fix to `reset_index` inplace in MultiIndex and other places
 - Pr #4899 Fix series inplace handling
 - PR #4889 Fix multi-index merging
+- PR #4922 Fix cudf::strings:split logic for many columns
 - PR #4949 Fix scatter, gather benchmark constructor call
 
 

cpp/CMakeLists.txt

@@ -630,6 +630,7 @@ add_library(cudf
             src/strings/sorting/sorting.cu
             src/strings/split/partition.cu
             src/strings/split/split.cu
+            src/strings/split/split_record.cu
             src/strings/strings_column_factories.cu
             src/strings/strings_column_view.cu
             src/strings/strings_scalar_factories.cpp

cpp/src/rolling/rolling.cu

@@ -368,7 +368,7 @@ struct rolling_window_launcher
           // The rows that represent null elements will be having negative values in gather map,
           // and that's why nullify_out_of_bounds/ignore_out_of_bounds is true.
           auto output_table = detail::gather(table_view{{input}}, output->view(), false, true, false, mr, stream);
-          return std::make_unique<cudf::column>(std::move(output_table->get_column(0)));;
+          output = std::make_unique<cudf::column>(std::move(output_table->get_column(0)));;
       }
 
       return output;