フリー問題

問題文

A developer needs a custom scoring function applied to a numeric column of a large DataFrame, and wants the function to receive many values at a time rather than one value per call. Which mechanism fits in PySpark?

選択肢

1. A plain Python user-defined function, because the engine automatically groups the rows of a partition into a single call when the function is registered with a return type that Spark can vectorize on its behalf.
2. A grouped map operation, applied per group of rows.
3. A pandas UDF that takes a series and returns a series of the same length.
4. A broadcast variable holding the function, which each executor then applies row by row.