Worked on the apache/spark repository to address business-value parity in PySpark DataFrame IO option handling. Focused on aligning Classic PySpark with Spark Connect by ensuring that Python None values in DataFrameReader and DataFrameWriter options are filtered out and treated as unset, rather than being forwarded as null to the JVM. Used Python and PySpark to implement this fix, expanding defensive test coverage across writer, V2 writer, streaming reader, and streaming writer APIs. Added cross-backend parity tests to verify consistent CSV null handling and updated migration documentation to guide users upgrading from PySpark 4.1 to 4.2.