Function make_sha2_udf 

pub fn make_sha2_udf() -> ScalarUDF

sha2(expr, bit_length) — Databricks/Spark-compatible SHA-2 returning a lowercase HEX STRING (Utf8), NOT raw bytes.

DataFusion’s built-in sha256() returns [DataType::Binary], so wrapping it in a string function (UPPER, CONCAT, …) fails at plan time. Databricks sha2(col, 256) instead yields a hex string, so the surrounding string ops are valid there. This UDF restores that contract: the digest bytes are identical to sha256(), rendered as the same lowercase-hex text Databricks produces, so surrogate keys stay byte-identical across engines.

bit_length selects the variant (224/256/384/512); 0 and any unsupported value map to 256. The second argument is accepted as Int64 so a bare SQL integer literal resolves without a cast; arg0 is accepted as Utf8.