From 5ce95f05eeae469f4d46516b3cf6fe19902623f6 Mon Sep 17 00:00:00 2001
From: David Wendt <45795991+davidwendt@users.noreply.github.com>
Date: Thu, 30 May 2024 09:24:58 -0400
Subject: [PATCH] Update interleave lists column for large strings (#15877)

Fixes the `compute_string_sizes_and_interleave_lists_fn` functor to use `column_device_view::element<string_view>()` method to access string row contents instead of using the strings offsets. This removes the need to add specific offsetalator logic to the logic.

Authors:
  - David Wendt (https://github.com/davidwendt)

Approvers:
  - Muhammad Haseeb (https://github.com/mhaseeb123)
  - Bradley Dice (https://github.com/bdice)

URL: https://github.com/rapidsai/cudf/pull/15877
---
 cpp/src/lists/interleave_columns.cu | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/cpp/src/lists/interleave_columns.cu b/cpp/src/lists/interleave_columns.cu
index be8fad62412..45ae3671d4e 100644
--- a/cpp/src/lists/interleave_columns.cu
+++ b/cpp/src/lists/interleave_columns.cu
@@ -166,8 +166,6 @@ struct compute_string_sizes_and_interleave_lists_fn {
       lists_col.child(lists_column_view::offsets_column_index).template data<size_type>() +
       lists_col.offset();
     auto const& str_col = lists_col.child(lists_column_view::child_column_index);
-    auto const str_offsets =
-      str_col.child(strings_column_view::offsets_column_index).template data<size_type>();
 
     // The range of indices of the strings within the source list.
     auto const start_str_idx = list_offsets[list_id];
@@ -181,13 +179,13 @@ struct compute_string_sizes_and_interleave_lists_fn {
     size_type write_idx = dst_list_offsets[idx];
 
     for (auto read_idx = start_str_idx; read_idx < end_str_idx; ++read_idx, ++write_idx) {
-      auto const offset        = str_offsets[read_idx];
-      auto const size          = str_offsets[read_idx + 1] - offset;
-      string_index_pair result = {nullptr, size};
-      if (str_col.is_valid(read_idx)) {
-        result.first = size > 0 ? str_col.template head<char>() + offset : "";
+      if (str_col.is_null(read_idx)) {
+        indices[write_idx] = string_index_pair{nullptr, 0};
+        continue;
       }
-      indices[write_idx] = result;
+      auto const d_str   = str_col.element<string_view>(read_idx);
+      indices[write_idx] = d_str.empty() ? string_index_pair{"", 0}
+                                         : string_index_pair{d_str.data(), d_str.size_bytes()};
     }
   }
 };