GH-45254: [C++][Acero] Fix the row offset truncation in row table mer…

…ge (#45255) ### Rationale for this change See #45254 ### What changes are included in this PR? First modify the test case to expose the suspecting bug. Then the fix in source. ### Are these changes tested? By existing tests. ### Are there any user-facing changes? None. * GitHub Issue: #45254 Authored-by: Rossi Sun <[email protected]> Signed-off-by: Rossi Sun <[email protected]>
apache · Jan 14, 2025 · ea47172 · ea47172
1 parent ef00568
commit ea47172
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 3 deletions.
diff --git a/cpp/src/arrow/acero/hash_join_node_test.cc b/cpp/src/arrow/acero/hash_join_node_test.cc
@@ -3370,8 +3370,10 @@ TEST(HashJoin, LARGE_MEMORY_TEST(BuildSideOver4GBVarLength)) {
   constexpr int value_no_match_length_min = 128;
   constexpr int value_no_match_length_max = 129;
   constexpr int value_match_length = 130;
+  // The value "DDD..." will be hashed to the partition over 4GB of the hash table.
+  // Matching at this area gives us more coverage.
   const auto value_match =
-      std::make_shared<StringScalar>(std::string(value_match_length, 'X'));
+      std::make_shared<StringScalar>(std::string(value_match_length, 'D'));
   constexpr int16_t num_rows_per_batch_left = 128;
   constexpr int16_t num_rows_per_batch_right = 4096;
   const int64_t num_batches_left = 8;

diff --git a/cpp/src/arrow/acero/swiss_join.cc b/cpp/src/arrow/acero/swiss_join.cc
@@ -439,11 +439,11 @@ Status RowArrayMerge::PrepareForMerge(RowArray* target,
     num_rows = 0;
     num_bytes = 0;
     for (size_t i = 0; i < sources.size(); ++i) {
-      target->rows_.mutable_offsets()[num_rows] = static_cast<uint32_t>(num_bytes);
+      target->rows_.mutable_offsets()[num_rows] = num_bytes;
       num_rows += sources[i]->rows_.length();
       num_bytes += sources[i]->rows_.offsets()[sources[i]->rows_.length()];
     }
-    target->rows_.mutable_offsets()[num_rows] = static_cast<uint32_t>(num_bytes);
+    target->rows_.mutable_offsets()[num_rows] = num_bytes;
   }
 
   return Status::OK();