jonasdedden opened a new issue, #51612:
URL: https://github.com/apache/arrow/issues/51612
### Describe the bug, including details regarding any error messages,
version, and platform.
Casting a `list_view` array to `list` (or `large_list_view` to `large_list`,
or across offset widths) returns wrong results whenever the views aren't laid
out like list offsets, e.g. when they're out of order or overlap.
```python
import pyarrow as pa
views = pa.ListViewArray.from_arrays(
pa.array([2, 0, 1], pa.int32()), # offsets
pa.array([2, 2, 1], pa.int32()), # sizes
pa.array([1, 2, 3, 4]),
)
print(views.to_pylist()) # [[3, 4], [1, 2], [2]]
print(views.cast(pa.list_(pa.int64())).to_pylist()) # [[], [1], []]
```
Expected: `[[3, 4], [1, 2], [2]]`.
The cast kernels are registered with the generic list-to-list cast
(`AddListCast<ListViewType, ListType>` and friends in
`cpp/src/arrow/compute/kernels/scalar_cast_nested.cc`, added in GH-39051).
`CastList::Exec` reuses the input's buffer 1 as the output's list offsets. For
a list view, buffer 1 holds the per-view offsets, and the sizes buffer is
ignored, so the result is only right when every `offsets[i] + sizes[i] ==
offsets[i + 1]`.
`ListArray::FromListView` / `LargeListArray::FromListView` already copy the
referenced values into a proper list layout, so the cast kernel could use that
(but see the related issue about `FromListView` and sliced inputs).
Reproduced on current `main` (C++ and a PyArrow dev build), Linux x86_64.
### Component(s)
C++, Python
--
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]
