[Data] Replace the unpack error when a V1 read finds no files

[jackylee-ch]

Description

FileBasedDatasource.__init__ unpacked the listing with zip(*expand_paths(...)), so an empty listing raised an internal error on every V1 file reader. Listing drops names starting with _ or ., so a Spark output directory holding only _SUCCESS reaches it without any typo.

ray.data.read_csv("/path/spark_output/")
# was: ValueError: not enough values to unpack (expected 2, got 0)
# now: ValueError: no files found under ['/path/spark_output']. Check the path
#      and any configured `partition_filter` or `file_extensions` filters.

The message is V2's, reused rather than invented. Missing paths are dropped during listing, so the old ignore_missing_paths branch could not tell absent paths from empty ones; both now report the same reason, with a note appended when the flag is set.

Related issues

None.

Additional information

Tests: empty directory, directory holding only _SUCCESS/a hidden file, all paths missing with the flag set. test_file_based_datasource.py 35 passed/3 skipped (was 31/3); csv/json/text/numpy/binary/webdataset suites pass; pre-commit clean; reverting the change fails all three new tests. Bazel/C++ not run locally.

No open PR touches this file. AI assistance (Claude) was used; I reviewed every changed line and ran the tests above.

[gemini-code-assist]

Code Review

This pull request refactors path expansion in FileBasedDatasource to raise a clear ValueError when no files are found, and adds unit tests to cover empty directories, excluded prefixes, and missing paths. The review feedback suggests refining the error message when ignore_missing_paths is enabled to avoid misleading users if the paths exist but are empty, along with updating the corresponding test assertions.

[gemini-code-assist]

When ignore_missing_paths is set to True, raising "None of the provided paths exist." can be misleading if the paths actually exist but are empty (or contain only excluded files, such as _SUCCESS or hidden files).

Consider using a more general and accurate error message that covers both cases (missing paths and empty directories).

Suggested change

	if ignore_missing_paths:
	raise ValueError(
	"None of the provided paths exist. "
	"The 'ignore_missing_paths' field is set to True."
	)
	if ignore_missing_paths:
	raise ValueError(
	f"No files found under {paths!r}. Check that the paths exist and "
	"are not empty (note that 'ignore_missing_paths' is set to True)."
	)

[gemini-code-assist]

Update the test assertion to match the improved error message.

Suggested change

	with pytest.raises(ValueError, match="None of the provided paths exist"):
	with pytest.raises(ValueError, match="No files found under"):

[jackylee-ch]

Confirmed, and the reason was wrong too — fixed in 961bc50.

Missing paths are dropped inside expand_paths, so an empty listing cannot tell absent paths from empty ones. read_csv("/tmp/empty_dir", ignore_missing_paths=True) reported "None of the provided paths exist" for a directory that does exist.

Both cases now report "no files found under ...", with a note appended when the flag is set. Test assertion updated. test_file_based_datasource.py 35 passed/3 skipped; csv/json/numpy suites pass.