[Data] Remove deprecated read_parquet_bulk #58969
Description
What should be done?
Remove the deprecated read_parquet_bulk API and its associated ParquetBulkDatasource class. The API was deprecated with a warning indicating removal after May 2025, and that deadline has passed.
Why?
The read_parquet_bulk function was deprecated in favor of read_parquet, which now handles all the same use cases. Keeping deprecated code around creates maintenance burden and confuses users.
The deprecation warning in the code states:
read_parquet_bulkis deprecated and will be removed after May 2025. Useread_parquetinstead.
Files to modify
1. **`python/ray/data/read_api.py`**
- Remove the `read_parquet_bulk` function (lines ~1227-1373)
- Remove any related imports (e.g., `ParquetBulkDatasource`)
2. **`python/ray/data/__init__.py`**
- Remove `read_parquet_bulk` from the import statement (line 69)
- Remove `read_parquet_bulk` from the `__all__` list (line 178)
3. **`python/ray/data/_internal/datasource/parquet_bulk_datasource.py`**
- Delete this entire file
4. **`python/ray/data/tests/test_parquet.py`**
- Remove tests that use `read_parquet_bulk` (search for `read_parquet_bulk` in the file)
5. **`doc/source/data/api/input_output.rst`**
- Remove `read_parquet_bulk` from the Parquet section (line 35)
6. **`python/ray/data/_internal/datasource/parquet_datasource.py`**
- Update the docstring that mentions `ParquetBulkDatasource` (line ~284)
7. Anything else this issue might've missed.
Verification
After making changes:
1. Run pre-commit to verify linting passes
2. Ensure the module still imports correctly: python -c "import ray.data"