Scala中rdd实现自定义分区

自定义分区

使用Scala做自定义分区，针对想要指定组合的分区情况
一种可以用sql语句简单快捷，另一种就是自定义分区，大家常常鲜

```java
import org.apache.spark.{Partitioner, SparkConf, SparkContext}
import scala.collection.mutable

object teacher_test2 {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setAppName("t_test2").setMaster("local[*]")
    val sc = new SparkContext(conf)
    val rdd = sc.textFile("C:\\Users\\Administrator\\Desktop\\teacher.txt")
    ***//数据：http://bigdata.baway.cn/shifo***
    val sub_teardd = rdd.map(line => {
      val fields = line.split("/")
      val sub = fields(2).split("[.]")(0)
      val teacher = fields(3)
      ((sub, teacher), 1)
    })
    val rdd2 = sub_teardd.reduceByKey(_+_)
    val subarr = rdd2.map(_._1._1).distinct().collect()

    *//实例化myPartition*
    val partition = new myPartition(subarr)
    val partitioned = rdd2.partitionBy(partition)
    val res = partitioned.mapPartitions(it => {
      it.toList.sortBy(_._2).reverse.take(2).toIterator
    })
    println(res.collect().toBuffer)
  }
}
class myPartition(subs:Array[String]) extends Partitioner{
  val rules = new mutable.HashMap[String,Int]()
  var i = 0
  for (sub<-subs){
    rules.put(sub,i)
    i += 1
  }
  ***//分区数量***
  override def numPartitions = subs.length

  ***//获取分区***
  override def getPartition(key: Any): Int = {
    val subsject = key.asInstanceOf[(String,String)]._1
    rules(subsject)
  }
}

**这是我敲的自定义分区，有不满意的请指点一二**