Spark Structured Streaming UDAF使用实例

class AreaClickUDAF extends UserDefinedAggregateFunction {
    // 输入数据的类型:  北京  String
    override def inputSchema: StructType = {
      StructType(StructField("city_name", StringType) :: Nil)
      //        StructType(Array(StructField("city_name", StringType)))
    }

    // 缓存的数据的类型: 北京->1000, 天津->5000  Map,  总的点击量  1000/?
    override def bufferSchema: StructType = {
      // MapType(StringType, LongType) 还需要标注 map的key的类型和value的类型
      StructType(StructField("city_count", MapType(StringType, LongType)) :: StructField("total_count", LongType) :: Nil)
    }

    // 输出的数据类型  "北京21.2%，天津13.2%，其他65.6%"  String
    override def dataType: DataType = StringType

    // 相同的输入是否应用有相同的输出.
    override def deterministic: Boolean = true

    // 给存储数据初始化
    override def initialize(buffer: MutableAggregationBuffer): Unit = {
      //初始化map缓存
      buffer(0) = Map[String, Long]()
      // 初始化总的点击量
      buffer(1) = 0L
    }

    // 分区内合并 Map[城市名, 点击量]
    override def update(buffer: MutableAggregationBuffer, input: Row): Unit = {
      // 首先拿到城市名, 然后把成名作为key去查看map中是否存在, 如果存在就把对应的值 +1, 如果不存在, 则直径0+1
      val cityName = input.getString(0)
      //        val map: collection.Map[String, Long] = buffer.getMap[String, Long](0)
      val map: Map[String, Long] = buffer.getAs[Map[String, Long]](0)
      buffer(0) = map + (cityName -> (map.getOrElse(cityName, 0L) + 1L))
      // 碰到一个城市, 则总的点击量要+1
      buffer(1) = buffer.getLong(1) + 1L
    }

    // 分区间的合并
    override def merge(buffer1: MutableAggregationBuffer, buffer2: Row): Unit = {
      val map1 = buffer1.getAs[Map[String, Long]](0)
      val map2 = buffer2.getAs[Map[String, Long]](0)

      // 把map1的键值对与map2中的累积, 最后赋值给buffer1
      buffer1(0) = map1.foldLeft(map2) {
        case (map, (k, v)) =>
          map + (k -> (map.getOrElse(k, 0L) + v))
      }

      buffer1(1) = buffer1.getLong(1) + buffer2.getLong(1)
    }

    // 最终的输出. "北京21.2%，天津13.2%，其他65.6%"
    override def evaluate(buffer: Row): Any = {
      val cityCountMap = buffer.getAs[Map[String, Long]](0)
      val totalCount = buffer.getLong(1)

      var citysRatio: List[CityRemark] = cityCountMap.toList.sortBy(-_._2).take(2).map {
        case (cityName, count) => {
          CityRemark(cityName, count.toDouble / totalCount)
        }
      }
      // 如果城市的个数超过2才显示其他
      if (cityCountMap.size > 2) {
        citysRatio = citysRatio :+ CityRemark("其他", citysRatio.foldLeft(1D)(_ - _.cityRatio))
      }
      citysRatio.mkString(", ")
    }
  }


  case class CityRemark(cityName: String, cityRatio: Double) {
    val formatter = new DecimalFormat("0.00%")
    override def toString: String = s"$cityName:${formatter.format(cityRatio)}"
  }


  def main(args: Array[String]): Unit = {
    val ss = SparkSession.builder().master("local[*]").appName("wula").enableHiveSupport().getOrCreate()
    val sc = ss.sparkContext
    sc.setLogLevel("warn")
    ss.sql("use testDemo01").show()

    //请使用spark sql 计算各个区域前三大热门商品，并备注上每个商品在主要城市中的分布比例，超过两个城市用其他显示。
    //用户行为表 时间 用户id 会话id 页面id 时间戳 搜索关键字 点击品类id 点击产品id 下单品类id 下单产品id 支付品类ids 支付产品ids 城市id
    //城市表     字段依次是：城市id 城市名字 地区
    //产品表：   产品id 产品名字 店铺类型
    ss.udf.register("cr", new AreaClickUDAF)
    val cvpDF1: DataFrame = ss.sql("select ci.*,pva.click_product_id,pi.product_name from product_info pi join user_visit_action pva on pi.product_id=pva.click_product_id join city_info ci on ci.city_id = pva.city_id where click_product_id>-1".stripMargin)
    cvpDF1.createOrReplaceTempView("tb1")
    //    cvpDF1.show()
    val cvpDF2: DataFrame = ss.sql("select tb1.area, tb1.product_name, count(*) click_count,cr(tb1.city_name) from tb1 group by area,product_name".stripMargin)
    cvpDF2.createOrReplaceTempView("tb2")
    ss.sql("select *,rank() over(partition by tb2.area order by tb2.click_count desc) rank from tb2 having rank <= 3".stripMargin).show()
    ss.stop()
  }