benibus commented on code in PR #35197:
URL: https://github.com/apache/arrow/pull/35197#discussion_r1175670935
##########
cpp/src/arrow/type.cc:
##########
@@ -1129,18 +1129,20 @@ class ChunkedArrayData : public ChunkedColumn {
// Return a vector of ChunkedColumns - one for each struct field.
// Unlike ChunkedArray::Flatten, this is zero-copy and doesn't merge
parent/child
// validity bitmaps.
-ChunkedColumnVector ChunkedColumn::Flatten() const {
+ChunkedColumnVector ChunkedColumn::FlattenZeroCopy() const {
DCHECK_EQ(type()->id(), Type::STRUCT);
ChunkedColumnVector columns(type()->num_fields());
for (int column_idx = 0; column_idx < type()->num_fields(); ++column_idx) {
const auto& child_type = type()->field(column_idx)->type();
ArrayDataVector chunks(num_chunks());
for (int chunk_idx = 0; chunk_idx < num_chunks(); ++chunk_idx) {
- const auto& child_data = chunk(chunk_idx)->child_data;
- DCHECK_EQ(columns.size(), child_data.size());
- DCHECK(child_type->Equals(child_data[column_idx]->type));
- chunks[chunk_idx] = child_data[column_idx];
+ const auto& parent = chunk(chunk_idx);
+ const auto& children = parent->child_data;
+ DCHECK_EQ(columns.size(), children.size());
+ auto child = children[column_idx]->Slice(parent->offset, parent->length);
Review Comment:
Yes (a bug that I introduced in https://github.com/apache/arrow/pull/34537).
It gives incorrect results when retrieving a nested child from an array sliced
from another array.
Although, now that I'm thinking about it, I'd imagine the same problem would
occur for the non-chunked `FieldPath::Get` variants as well. I'll need to test
it and create a separate issue.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
