[Python] pyarrow.dataset.dataset does not accept RecordBatchReader as source

[sugibuchi]

Describe the bug, including details regarding any error messages, version, and platform.

The documentation of pyarrow.dataset.dataset says this function accepts RecordBatchReader as source.

(List of) batches or tables, iterable of batches, or RecordBatchReader:
Create an InMemoryDataset. If an iterable or empty list is given, a schema must also be given. If an iterable or RecordBatchReader is given, the resulting dataset can only be scanned once; further attempts will raise an error.
https://arrow.apache.org/docs/python/generated/pyarrow.dataset.dataset.html#pyarrow-dataset-dataset

However, pyarrow.dataset.dataset throws TypeError when we call this function with RecordBatchReader as source.

Environment

• OS: Ubuntu 22.04
• Python: Python 3.9.18
• PyArrow: 13.0.0

POC

import pyarrow as pa
import pyarrow.dataset as ds

table = pa.Table.from_pydict({
    "col_1": list(range(0, 10000)),
    "col_2": [f"v{v}" for v in range(0, 10000)]
})

batches = t.to_batches(max_chunksize=100)

ds.dataset(batches) # -> <pyarrow._dataset.InMemoryDataset at ...>

batch_reader = pa.RecordBatchReader.from_batches(table.schema, batches)
ds.dataset(batch_reader) # -> Fail!

The last line fails with the following error.

File /opt/conda/lib/python3.9/site-packages/pyarrow/dataset.py:793, in dataset(source, schema, format, filesystem, partitioning, partition_base_dir, exclude_invalid_files, ignore_prefixes)
    791     return _in_memory_dataset(source, **kwargs)
    792 else:
--> 793     raise TypeError(
    794         'Expected a path-like, list of path-likes or a list of Datasets '
    795         'instead of the given type: {}'.format(type(source).__name__)
    796     )

TypeError: Expected a path-like, list of path-likes or a list of Datasets instead of the given type: RecordBatchReader

Component(s)

Python

[mapleFU]

        elif all(isinstance(elem, (pa.RecordBatch, pa.Table))
                 for elem in source):
            return _in_memory_dataset(source, **kwargs)

Seems from code it should be record-batch or table, maybe first add a casting as a workaround?

[kou]

@lidavidm It seems that this is related to #10070 / #28047.
Should we update the document?

[lidavidm]

Yes, we should update the docs

[kou]

OK.
@sugibuchi Do you want to work on this?