Question

我正在尝试编写通用的add方法，该方法可以采用任何参数类型并返回该类型的结果

 def addExactUDF[T](x: T, y: T)(implicit num: Numeric[T]): T = {
    import num._
    x + y
  }

  def addExact(value1: Column, value2: Column, dataType: String): Column =
    dataType match {
      case "Int" => expr(s"addExactUDF(cast($value1 AS INT), cast($value2 AS INT))")
      case "Double" => expr(s"addExactUDF(cast($value1 AS DOUBLE), cast($value2 AS DOUBLE))")
    }

现在当我尝试注册UDF时

object FilterFunctionsUtil extends MathFunctionsNameSpace with StringFunctionsNameSpace {
  lazy val registerAsSparkUdf: UserDefinedFunction = {
    val sqlContext = SparkSessionFactory.getSparkSession(Map(), Nil).sqlContext

    sqlContext.udf.register("addExactUDF", addExactUDF _)
    sqlContext.udf.register("subtractExactUDF", subtractExactUDF _)
  }
}

这让我出错

   Error:(36, 44) could not find implicit value for parameter num: Numeric[Nothing]
        sqlContext.udf.register("addExactUDF", addExactUDF _)
Error:(36, 44) not enough arguments for method addExactUDF: (implicit num: Numeric[Nothing])Nothing.
Unspecified value parameter num.
    sqlContext.udf.register("addExactUDF", addExactUDF _)

我如何添加隐式函数以使其正常工作

Answer 1

在注册UDF时指定隐式类型。您可能需要根据需要为每种类型分别注册UDF。

  import sqlContext.implicits._
  val df = sc.parallelize(Seq((1,2),(3,4))).toDF("val1","val2")

  df.createOrReplaceTempView("tempTable")
  sqlContext.cacheTable("tempTable")

  sqlContext.udf.register("addExtractIntUDF",addExtractUDF[Int] _)

  sqlContext.sql("select addExtractIntUDF(val1,val2) from temptable").show(false)

结果

+ --------------------------------------------- +
| UDF：addExtractIntUDF（val1，val2）|
+ --------------------------------------------- +
| 3的| 3的 | | | | | | 7 |
+ -------------------------------------------- +

Spark UDF找不到参数num的隐式值：Numeric [Nothing]

1 个答案: