adamreeve opened a new issue, #55447:
URL: https://github.com/apache/arrow/issues/55447
### Describe the enhancement requested
Trying to create a dataset as a union of filtered datasets currently raises
an error:
```python
from pathlib import Path
import tempfile
import pyarrow as pa
import pyarrow.compute as pc
import pyarrow.dataset as ds
with tempfile.TemporaryDirectory() as temp_dir:
temp_path = Path(temp_dir)
t1 = pa.table({"x": [1, 1, 2, 2], "y": [1, 2, 3, 4]})
ds.write_dataset(t1, temp_path / "t1", format="parquet")
ds1 = ds.dataset(temp_path / "t1", format="parquet")
t2 = pa.table({"x": [1, 1, 2, 2], "y": [5, 6, 7, 8]})
ds.write_dataset(t2, temp_path / "t2", format="parquet")
ds2 = ds.dataset(temp_path / "t2", format="parquet")
ds1f = ds1.filter(pc.field("x") > 1)
ds2f = ds2.filter(pc.field("x") < 2)
union = ds.dataset([ds1f, ds2f])
```
Raises:
```
ValueError: Creating an UnionDataset from filtered or projected Datasets is
currently not supported. Union the unfiltered datasets and apply the filter to
the resulting union.
```
The suggested workaround is only valid if the filter applied to each dataset
is the same. If the filters are different, or not all datasets are filtered,
then there's currently no way to create a dataset that represents the union of
these datasets.
This limitation seems to be present because the filter is only stored on the
Python Dataset object, but creating a union dataset happens in C++.
### 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]
