spark实现将相同用户（key）所有item列表聚合

数据：用户id，itemid，分数

代码：

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

object userwatchlist {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf()
    //conf.setMaster("local")
    conf.setAppName("userwatchlist test")

    val sc = new SparkContext(conf)
    val input_path = sc.textFile("/train_new.data")
    val output_path = "/userwatchlist_output"

    val data = input_path.filter{ x =>
      val fields = x.split("\t")
      fields(2).toDouble > 2
    }.map { x =>
      val fields = x.split("\t")
      (fields(0).toString, (fields(1).toString, fields(2).toString))
    }.groupByKey(10).map { x =>
      val userid = x._1
      val item_score_tuple_list = x._2
      val tmp_arr = item_score_tuple_list.toArray.sortWith(_._2>_._2)
      var watchlen = tmp_arr.length
      if (watchlen > 5){
        watchlen = 5
      }

      val strbuf = new StringBuilder
      for (i <- 0 until watchlen) {
        strbuf ++= tmp_arr(i)._1
        strbuf.append(":")
        strbuf ++= tmp_arr(i)._2
        strbuf.append(" ")
      }
      userid + "\t" + strbuf

    }

    data.saveAsTextFile(output_path)

  }
}

结果：

将相同用户（key）所有item列表放在了一行