GH-47728: [Python] Check the source argument in parquet.read_table

[rmnskb]

Rationale for this change

See #47728. Check source argument in pyarrow.parquet.read_table if pyarrow.dataset is not available.

What changes are included in this PR?

Check the source argument, raise ValueError if the source argument is either a list of .parquet files or a directory.

Are these changes tested?

Yes

Are there any user-facing changes?

No

In case if the source argument is a directory, I decided not to check it directly, but to catch the exceptions coming from the fs.open_input_file, since it already checks for it, and add extra exception on top of the stack that explains the actual reason.

• GitHub Issue: [Python] Check the source argument in parquet.read_table #47728

[raulcd]

Thanks for working on this!

[raulcd]

The test failure test_read_table_without_dataset failing on the CI jobs is related. It seems a test that was using a MockDataset with a non-existing file is failing with this change, we might have to revisit that test.

[rmnskb]

The test failure test_read_table_without_dataset failing on the CI jobs is related. It seems a test that was using a MockDataset with a non-existing file is failing with this change, we might have to revisit that test.

Yep, you're right, I've updated it with the new ValueError.

[AlenkaF]

Thank you for another contribution @rmnskb!
LGTM. I just added minor styling suggestions.

[AlenkaF]

Suggested change

	def test_read_table_raises_value_error_when_ds_is_unavailable(
	monkeypatch, source):
	def test_read_table_raises_value_error_when_ds_is_unavailable(monkeypatch, source):