一、自定義分區

　　1.概述

　　　　默認的是Hash的分區策略，這點和Hadoop是類似的，具體的分區介紹，參見：https://blog.csdn.net/high2011/article/details/68491115

　　2.實現

package cn.itcast.spark.day3

import java.net.URL
import org.apache.spark.{HashPartitioner, Partitioner, SparkConf, SparkContext}
import scala.collection.mutable

/**
  * Created by root on 2016/5/18.
  
 
*/
object UrlCountPartition {

  def main(args: Array[String]) {

    val conf = new SparkConf().setAppName("UrlCountPartition").setMaster("local[2]")
    val sc = new SparkContext(conf)

    //rdd1將數據切分，元組中放的是（URL， 1）
    val rdd1 = sc.textFile("c://itcast.log").map(line => {
      val f = line.split("\t")
      (f(
 
1), 1)
    })
    val rdd2 = rdd1.reduceByKey(_ + _)

    val rdd3 = rdd2.map(t => {
      val url = t._1
      val host = new URL(url).getHost
      (host, (url, t._2))
    })
    val ints = rdd3.map(_._1).distinct().collect()
    val hostParitioner = new HostParitioner(ints)
//    val rdd4 = rdd3.partitionBy(new HashPartitioner(ints.length))
 


    val rdd4 = rdd3.partitionBy(hostParitioner).mapPartitions(it => {
      it.toList.sortBy(_._2._2).reverse.take(2).iterator
    })
    rdd4.saveAsTextFile("c://out4")
    //println(rdd4.collect().toBuffer)
    sc.stop()
  }
}

/**
  * 決定了數據到哪個分區裏面
  * @param ins
  */
class HostParitioner(ins: Array[String]) extends Partitioner {

  val parMap = new mutable.HashMap[String, Int]()
  var count = 0
  for(i <- ins){
    parMap += (i -> count)
    count += 1
  }

  override def numPartitions: Int = ins.length

  override def getPartition(key: Any): Int = {
    parMap.getOrElse(key.toString, 0)
  }
}

　　// 與Hadoop相通，不再贅述

二、自定義排序

　　基本上就是結合之前的隱式轉換了：（這裏使用樣例類可以不用new就能得到實例，另外也可以用於模式匹配）

package cn.itcast.spark.day3

import org.apache.spark.{SparkConf, SparkContext}


object OrderContext {
  implicit val girlOrdering  = new Ordering[Girl] {
    override def compare(x: Girl, y: Girl): Int = {
      if(x.faceValue > y.faceValue) 1
      else if (x.faceValue == y.faceValue) {
        if(x.age > y.age) -1 else 1
      } else -1
    }
  }
}


/**
  * Created by root on 2016/5/18.
  */
//sort =>規則 先按faveValue，比較年齡
//name,faveValue,age


object CustomSort {

  def main(args: Array[String]) {
    val conf = new SparkConf().setAppName("CustomSort").setMaster("local[2]")
    val sc = new SparkContext(conf)
    val rdd1 = sc.parallelize(List(("yuihatano", 90, 28, 1), ("angelababy", 90, 27, 2),("JuJingYi", 95, 22, 3)))
    import OrderContext._
    val rdd2 = rdd1.sortBy(x => Girl(x._2, x._3), false)
    println(rdd2.collect().toBuffer)
    sc.stop()
  }

}

/**
  * 第一種方式
  * @param faceValue
  * @param age

case class Girl(val faceValue: Int, val age: Int) extends Ordered[Girl] with Serializable {
  override def compare(that: Girl): Int = {
    if(this.faceValue == that.faceValue) {
      that.age - this.age
    } else {
      this.faceValue -that.faceValue
    }
  }
}
  */

/**
  * 第二種，通過隱式轉換完成排序
  * @param faceValue
  * @param age
  */
case class Girl(faceValue: Int, age: Int) extends Serializable

大數據入門第二十二天——spark（三）自定義分區、排序與查找