spark 例子groupByKey分组计算2

---

例子描述：

大概意思为，统计用户使用app的次数排名

原始数据：

000041b232,张三,FC:1A:11:5C:58:34,F8:E7:1E:1E:62:20,15097003,,2016/6/8 17:10,2016/6/8 17:10,690,6218,11=0|12=200,2016/7/5 11:11

000041b232,张三,FC:1A:11:5C:58:34,F8:E7:1E:1E:69:C0,15026002,,2016/6/8 17:10,2016/6/8 17:10,690,6218,11=0|12=200,2016/7/5 11:11

000041b232,张三,FC:1A:11:5C:58:34,F8:E7:1E:1E:62:20,15026002,,2016/6/8 17:10,2016/6/8 17:10,690,6218,11=0|12=200,2016/7/5 11:11

000041b744,张三,FC:1A:11:5C:58:34,F8:E7:1E:1E:62:20,15026002,,2016/6/8 17:10,2016/6/8 17:10,719,4174,6=2016-06-23 08:50:00|7=,2016/7/5 11:11

000041b22f,李四,FC:1A:11:5C:58:34,F8:E7:1E:1E:62:20,15097002,,2016/6/8 17:10,2016/6/8 17:10,856,367,7=,2016/7/5 11:11

000041b1bc,李四,FC:1A:11:5C:58:34,F8:E7:1E:1E:62:20,15026002,,2016/6/8 17:10,2016/6/8 17:10,937,2964,3=北京|4=上海,2016/7/5 11:11

000041cf18,赵六,7C:1D:D9:F4:BE:E0,F8:E7:1E:1E:62:20,15097002,,2016/6/8 17:10,2016/6/8 17:10,665,2669,5=2016-06-22 00:00:00,2016/7/5 11:11

000041b1bc,孙七,7C:1D:D9:F4:BE:E0,38:FF:36:2E:5B:A0,9003000,,2016/6/8 17:10,2016/6/8 17:10,530,245,,2016/7/5 11:11

000041b8f1,王五,FC:1A:11:5C:58:34,38:FF:36:2E:5B:A0,9007000,,2016/6/8 17:11,2016/6/8 17:11,626,6886,,2016/7/5 11:11

000041b8f1,周八,FC:1A:11:5C:58:34,38:FF:36:2E:5B:A0,16500000,,2016/6/8 17:11,2016/6/8 17:11,2532,646,,2016/7/5 11:11

000041966a,李四,FC:1A:11:5C:58:34,38:FF:36:2E:5B:A0,16501000,,2016/6/8 17:11,2016/6/8 17:11,690,454,,2016/7/5 11:11

000041966a,李四,FC:1A:11:5C:58:34,38:FF:36:2E:5B:A0,16501000,,2016/6/8 17:11,2016/6/8 17:11,690,454,,2016/7/5 11:11

结果数据：

周八,人人贷:1

孙七,支付宝:1

赵六,途牛机票:1

王五,快钱:1|天弘基金:1

李四,红岭创投:2|携程机票:1|携程酒店:1|途牛机票:1

张三,途牛酒店:5|携程机票:3

---

代码片段：

cxRDD0.map {

  lines =>

    val line = lines.split(",")//逗号分隔数据

    //想办法将数据拼成(数据,1)的映射,并且这个地方的数据要相同,可以理解取为用户,APPID,然后当成K,写个数字1当成V，这里使用的字典关联去取的数据

    (s"""${line((data_location.getOrElse("USR_NBR", "").toInt))},${buss_location.getOrElse(line((data_location.getOrElse("BUS_ID", "").toInt)), "").split(",", -1)(0)}""", 1)

}.reduceByKey(_ + _).map {//分组

  lines =>

    //将分组后的数据,以用户为K,其他为V拼成映射,便于后续分组

    (s"${lines._1.split(",")(0)}", s"${lines._1.split(",")(1)},${lines._2}")

}.groupByKey().map {//分组

  case (k, v) =>

    //对APPID数量 V 进行排序

    val app = v.map {

      x =>

        val a = x.split(",")

        //拆分APPID 与 数量,这里传递给下面的类型为映射

        (a(0), a(1))

        //使用sortWith对映射的第二位数字进行排序,需要转换成INT,因为传递过来都是字符

    }.toSeq.sortWith(_._2.toInt > _._2.toInt).map {

      app =>

        //格式化输出

        //V:V

        s"${app._1}:${app._2}"

    }

    //格式化输出

    //K,V

    //K,V1|V2......

    s"$k,${app.mkString("|")}"

}.foreach(println)