spark中reduceByKey、groupByKey、combineByKey的区别

reduceByKey

reduceByKey用于对每个key对应的多个value进行merge操作，最重要的是它能够在本地先进行merge操作，并且merge操作可以通过函数自定义；

groupByKey

groupByKey也是对每个key进行操作，但只生成一个sequence，groupByKey本身不能自定义函数，需要先生成RDD，然后才能对此RDD通过map进行自定义函数操作

比较发现，使用groupByKey时，不会进行局部merge，会导致集群节点之间的开销很大，导致传输延时。

combineByKey

import org.apache.log4j.{Level, Logger}
import org.apache.spark.rdd.RDD
import org.apache.spark.util.LongAccumulator
import org.apache.spark.{SparkConf, SparkContext}

import scala.collection.mutable

object Test{
	def main(args: Array[String]): Unit = {
		Logger.getLogger("org.apache.hadoop").setLevel(Level.WARN)
		Logger.getLogger("org.apache.spark").setLevel(Level.WARN)
		Logger.getLogger("org.spark-project").setLevel(Level.WARN)
		val conf = new SparkConf()
				.setAppName(s"${Test.getClass.getSimpleName}")
				.setMaster("local[*]")
		val sc = new SparkContext(conf)
		val lines = sc.textFile("file:///D:/data/people.txt")
		val man:LongAccumulator = lines.sparkContext.longAccumulator("man")
		val woman:LongAccumulator = lines.sparkContext.longAccumulator("woman")
		val peoMapRDD:RDD[(String,String)] = lines.map(line => {
			val files = line.split("\\s+")
			(files(1),files(2))
		})
		peopleOps(sc,peoMapRDD)
		sc.stop()
	}

	def peopleOps(sc: SparkContext,peoMapRDD:RDD[(String,String)]):Unit = {
		def createCombiner(info:String):mutable.TreeSet[String] = {
			val ple = mutable.TreeSet[String]()
			ple.add(info)
			ple
		}
		def mergeValue(ple:mutable.TreeSet[String],info:String):mutable.TreeSet[String] = {
			ple.add(info)
			ple
		}
		def mergeCombiners(ple1:mutable.TreeSet[String],ple2:mutable.TreeSet[String]):mutable.TreeSet[String] = {
			val ple = mutable.TreeSet[String]()(new Ordering[String](){
				override def compare(x: String, y: String): Int = {
					y.compareTo(x)
				}
			})
			for(info <- ple1.++:(ple2)){
				ple.add(info)
			}
			ple
		}
		val ret = peoMapRDD.combineByKey(createCombiner,mergeValue,mergeCombiners)
		ret.foreach(println)
	}

}

people.txt数据：
1 F 170
2 M 178
3 M 174
4 F 165