【Spark征服之路-3.6-Spark-SQL核心编程（五）】

自定义函数：

UDF：

val sparkConf = new SparkConf().setMaster("local[*]").setAppName("SQLDemo")
//创建SparkSession对象
val spark :SparkSession = SparkSession.builder().config(sparkConf).getOrCreate()

import spark.implicits._
//读取json文件
val df : DataFrame = spark.read.json("Spark-SQL/input/user.json")

spark.udf.register("addName",(x:String)=>"Name:"+x)

df.createOrReplaceTempView("people")
spark.sql("select addName(username),age from people").show()

spark.stop()

UDAF（自定义聚合函数）

强类型的 Dataset 和弱类型的 DataFrame 都提供了相关的聚合函数， 如 count()，

countDistinct()，avg()，max()，min()。除此之外，用户可以设定自己的自定义聚合函数。Spark3.0之前我们使用的是UserDefinedAggregateFunction作为自定义聚合函数，从 Spark3.0 版本后可以统一采用强类型聚合函数 Aggregator

实验需求：计算平均工资

实现方式一：RDD

val sparkconf: SparkConf = new SparkConf().setAppName("app").setMaster("local[*]")
val sc: SparkContext = new SparkContext(conf)
val resRDD: (Int, Int) = sc.makeRDD(List(("zhangsan", 20), ("lisi", 30), ("wangwu",40))).map {
case (name, salary) => {
    (salary, 1)
  }
}.reduce {
  (t1, t2) => {
    (t1._1 + t2._1, t1._2 + t2._2)
  }
}
println(resRDD._1/resRDD._2)
// 关闭连接
sc.stop()

实现方式二：弱类型UDAF

class MyAverageUDAF extends UserDefinedAggregateFunction{
def inputSchema: StructType =
StructType(Array(StructField("salary",IntegerType)))
// 聚合函数缓冲区中值的数据类型(salary,count)
  def bufferSchema: StructType = {

StructType(Array(StructField("sum",LongType),StructField("count",LongType)))
  }
// 函数返回值的数据类型
  def dataType: DataType = DoubleType
// 稳定性：对于相同的输入是否一直返回相同的输出。
  def deterministic: Boolean = true
  // 函数缓冲区初始化
  def initialize(buffer: MutableAggregationBuffer): Unit = {
// 存薪资的总和
    buffer(0) = 0L
// 存薪资的个数
    buffer(1) = 0L
}
// 更新缓冲区中的数据
  def update(buffer: MutableAggregationBuffer,input: Row): Unit = {
if (!input.isNullAt(0)) {
      buffer(0) = buffer.getLong(0) + input.getInt(0)
      buffer(1) = buffer.getLong(1) + 1
}
  }
// 合并缓冲区
  def merge(buffer1: MutableAggregationBuffer,buffer2: Row): Unit = {
    buffer1(0) = buffer1.getLong(0) + buffer2.getLong(0)
    buffer1(1) = buffer1.getLong(1) + buffer2.getLong(1)
  }
// 计算最终结果
  def evaluate(buffer: Row): Double = buffer.getLong(0).toDouble /
    buffer.getLong(1)
}

val sparkconf: SparkConf = new SparkConf().setAppName("app").setMaster("local[*]")
val spark:SparkSession = SparkSession.builder().config(conf).getOrCreate()

import spark.implicits._
val res :RDD[(String,Int)]= spark.sparkContext.makeRDD(List(("zhangsan", 20), ("lisi", 30), ("wangwu",40)))

val df :DataFrame = res.toDF("name","salary")
df.createOrReplaceTempView("user")
var myAverage = new MyAverageUDAF
//在 spark 中注册聚合函数
spark.udf.register("avgSalary",myAverage)
spark.sql("select avgSalary(salary) from user").show()

// 关闭连接
spark.stop()

实现方式三：强类型UDAF

case class Buff(var sum:Long,var cnt:Long)
class MyAverageUDAF extends Aggregator[Long,Buff,Double]{
override def zero: Buff = Buff(0,0)
override def reduce(b: Buff, a: Long): Buff = {
    b.sum += a
    b.cnt += 1
b
  }
override def merge(b1: Buff, b2: Buff): Buff = {
    b1.sum += b2.sum
    b1.cnt += b2.cnt
    b1
  }
override def finish(reduction: Buff): Double = {
    reduction.sum.toDouble/reduction.cnt
  }
override def bufferEncoder: Encoder[Buff] = Encoders.product
  override def outputEncoder: Encoder[Double] = Encoders.scalaDouble

}

val sparkconf: SparkConf = new SparkConf().setAppName("app").setMaster("local[*]")
val spark:SparkSession = SparkSession.builder().config(conf).getOrCreate()

import spark.implicits._
val res :RDD[(String,Int)]= spark.sparkContext.makeRDD(List(("zhangsan", 20), ("lisi", 30), ("wangwu",40)))

val df :DataFrame = res.toDF("name","salary")
df.createOrReplaceTempView("user")
var myAverage = new MyAverageUDAF
//在 spark 中注册聚合函数
spark.udf.register("avgSalary",functions.udaf(myAverage))
spark.sql("select avgSalary(salary) from user").show()

// 关闭连接
spark.stop()