1 import org.apache.spark.sql.{Row, SparkSession}
 2 import org.apache.spark.sql.expressions.{MutableAggregationBuffer, UserDefinedAggregateFunction}
 3 import org.apache.spark.sql.types._
 4 
 5 /**
 6   * Created by zhen on 2018/11/26.
 7   */
 8 object AverageUserDefinedAggregateFunction extends UserDefinedAggregateFunction{
 
 9   //聚合函式輸入資料結構
10   override def inputSchema:StructType = StructType(StructField("input", LongType) :: Nil)
11 
12   //快取區資料結構
13   override def bufferSchema: StructType = StructType(StructField("sum", LongType) :: StructField("count", LongType) :: Nil)
14 
15   //結果資料結構
16   override def dataType : DataType = DoubleType
 
17 
18   // 是否具有唯一性
19   override def deterministic : Boolean = true
20 
21   //初始化
22   override def initialize(buffer : MutableAggregationBuffer) : Unit = {
23     buffer(0) = 0L
24     buffer(1) = 0L
25   }
26 
27   //資料處理 : 必寫，其它方法可選，使用預設
28   override def update(buffer: MutableAggregationBuffer, input: Row): Unit = {
 
29     if(input.isNullAt(0)) return
30     buffer(0) = buffer.getLong(0) + input.getLong(0) //求和
31     buffer(1) = buffer.getLong(1) + 1 //計數
32   }
33 
34   //合併
35   override def merge(bufferLeft: MutableAggregationBuffer, bufferRight: Row): Unit ={
36     bufferLeft(0) = bufferLeft.getLong(0) + bufferRight.getLong(0)
37     bufferLeft(1) = bufferLeft.getLong(1) + bufferRight.getLong(1)
38   }
39 
40   //計算結果
41   override def evaluate(buffer: Row): Any  = buffer.getLong(0).toDouble / buffer.getLong(1)
42 }
43 object SparkUdaf {
44   def main(args: Array[String]) {
45     val spark = SparkSession
46       .builder()
47       .appName("udaf")
48       .master("local[2]")
49       .getOrCreate()
50 
51     spark.read.json("E:/BDS/newsparkml/src/udaf.json").createOrReplaceTempView("user")
52     spark.udf.register("average", AverageUserDefinedAggregateFunction)
53     spark.sql("select count(*) count,average(age) avg_age from user").show()
54 
55   }
56 }

結果：