用spark实现count(distinct fieldname)形式的聚合

举个例子,比如要统计用户的总访问次数和去除访问同一个URL之后的总访问次数,随便造了几条样例数据(四个字段:id,name,vtm,url,vtm字段本例没用,不用管)如下:

id1,user1,2,http://www.hupu.com
id1,user1,2,http://www.hupu.com
id1,user1,3,http://www.hupu.com
id1,user1,100,http://www.hupu.com
id2,user2,2,http://www.hupu.com
id2,user2,1,http://www.hupu.com
id2,user2,50,http://www.hupu.com
id2,user2,2,http://touzhu.hupu.com

根据这个数据集,我们可以写hql 实现:select id,name, count(0) as ct,count(distinct url) as urlcount from table group by id,name.

得出结果应该是:

id1,user1,4,1

id2,user2,4,2

下面用spark实现这个聚合功能:

import org.apache.spark.SparkContext._
import org.apache.spark._

/**
 * Created by xiaojun on 2015/3/9.
 */
object SparkDemo2 {
  def main(args: Array[String]) {

    case class User(id: String, name: String, vtm: String, url: String)
    //val rowkey = (new RowKey).evaluate(_)
    val HADOOP_USER = "hdfs"
    // 设置访问spark使用的用户名
    System.setProperty("user.name", HADOOP_USER);
    // 设置访问hadoop使用的用户名
    System.setProperty("HADOOP_USER_NAME", HADOOP_USER);

    val conf = new SparkConf().setAppName("wordcount").setMaster("local").setExecutorEnv("HADOOP_USER_NAME", HADOOP_USER)
    val sc = new SparkContext(conf)
    val data = sc.textFile("test.txt")
    val rdd1 = data.map(line => {
      val r = line.split(",")
      User(r(0), r(1), r(2), r(3))
    })
    val rdd2 = rdd1.map(r => ((r.id, r.name), r))

    val seqOp = (a: (Int, List[String]), b: User) => a match {
      case (0, List()) => (1, List(b.url))
      case _ => (a._1 + 1, b.url :: a._2)
    }

    val combOp = (a: (Int, List[String]), b: (Int, List[String])) => {
      (a._1 + b._1, a._2 ::: b._2)
    }

    println("-----------------------------------------")
    val rdd3 = rdd2.aggregateByKey((0, List[String]()))(seqOp, combOp).map(a => {
      (a._1, a._2._1, a._2._2.distinct.length)
    })
    rdd3.collect.foreach(println)
    println("-----------------------------------------")
    sc.stop()
  }

}

输出结果如下:

((id1,user1),4,1)
((id2,user2),4,2)
与HQL输出一致.