SparkSQL UDAF

package com.it.baizhan.scalacode.sparksql

import org.apache.spark.sql.{Row, SparkSession}
import org.apache.spark.sql.expressions.{MutableAggregationBuffer, UserDefinedAggregateFunction}
import org.apache.spark.sql.types.{DataType, DataTypes, StructField, StructType}

class MyUDAF extends UserDefinedAggregateFunction {
    //调用UDAF函数时，传参的类型
    override def inputSchema: StructType = StructType(List[StructField](
      StructField("xx",DataTypes.StringType)
    ))

    //设置 在计算过程中，更新的数据类型
    override def bufferSchema: StructType = StructType(List[StructField](
      StructField("xx",DataTypes.IntegerType)
    ))

    //指定调用函数最后返回数据类型
    override def dataType: DataType = DataTypes.IntegerType

    //多次运行，结果顺序保持一致
    override def deterministic: Boolean = true

    // 作用在map和reduce两侧，给每个分区内的每个分组的数据做初始值
    override def initialize(buffer: MutableAggregationBuffer): Unit = buffer.update(0,0)
    //作用在map端每个分区内的每个分组上
    override def update(buffer: MutableAggregationBuffer, input: Row): Unit = buffer.update(0,buffer.getInt(0)+1)
    //作用在reduce端，每个分区的每个分组上，对map的结果做聚合
    override def merge(buffer1: MutableAggregationBuffer, buffer2: Row): Unit = buffer1.update(0,buffer1.getInt(0)+buffer2.getInt(0))

    //调用函数最后返回的数据结果
    override def evaluate(buffer: Row): Any = buffer.getInt(0)
}

object SparkSQLUDAF {
  def main(args: Array[String]): Unit = {
    val session = SparkSession.builder().master("local").appName("test").getOrCreate()
    val nameList = List[String]("zhangsan","lisi","zhangsan","zhangsan","zhangsan","lisi","wangwu","wangwu","lisi","maliu")
    import session.implicits._
    val frame = nameList.toDF("name")
    frame.createTempView("infos")

    /**
      * 可以自己定义聚合函数实现 多行数据对应一个结果的功能。例如：自定义UDAF函数实现一个count功能
      */

    session.udf.register("namecount",new MyUDAF())

//    session.udf.register("namecount",new UserDefinedAggregateFunction {
//      //调用UDAF函数时，传参的类型
//      override def inputSchema: StructType = StructType(List[StructField](
//        StructField("xx",DataTypes.StringType)
//      ))
//
//      //设置 在计算过程中，更新的数据类型
//      override def bufferSchema: StructType = StructType(List[StructField](
//        StructField("xx",DataTypes.IntegerType)
//      ))
//
//      //指定调用函数最后返回数据类型
//      override def dataType: DataType = DataTypes.IntegerType
//
//      //多次运行，结果顺序保持一致
//      override def deterministic: Boolean = true
//
//      // 作用在map和reduce两侧，给每个分区内的每个分组的数据做初始值
//      override def initialize(buffer: MutableAggregationBuffer): Unit = buffer.update(0,0)
//      //作用在map端每个分区内的每个分组上
//      override def update(buffer: MutableAggregationBuffer, input: Row): Unit = buffer.update(0,buffer.getInt(0)+1)
//      //作用在reduce端，每个分区的每个分组上，对map的结果做聚合
//      override def merge(buffer1: MutableAggregationBuffer, buffer2: Row): Unit = buffer1.update(0,buffer1.getInt(0)+buffer2.getInt(0))
//
//      //调用函数最后返回的数据结果
//      override def evaluate(buffer: Row): Any = buffer.getInt(0)
//    })

    session.sql(
      """
        | select name,namecount(name) as totalCount from infos group by name
      """.stripMargin).show()
  }

}