At the moment, Spark does not support window functions (like `lag` or `head`) on a stream. It would be great if you could add support for those type of operation.