scala - Which one yields better performance in spark data frame?

Question

Which one should be faster here ? Or both have same performance level?

-) Normal Function

  private def fun(lc: Column): Column = lc -  lag(lc, offset = 10, defaultValue = 0).over(Window.partitionBy("colN").orderBy())

-) UDF

val funUdf = udf (fun_)

1) df.withColumn("abc", fun("col1"))

2) df.withColumn("abc", funUdf ("col1"))

Which one will perform faster ?

Answer 1

等待大神答复