Spark核心编程-分组取topN

案例需求

对每个班级内的学生成绩,取出前3名。(分组取topN)


输入测试数据(以“ ”以做分割符)

class1 90
class2 56
class1 87
class1 76
class2 88
class1 95
class1 74
class2 87
class2 67
class2 77
class1 98
class2 96 



实现如下:

1 、scala的版本

package com.spark.core

import org.apache.spark.SparkConf
import org.apache.spark.SparkContext
import scala.collection.mutable.ArrayBuffer
import scala.util.control.Breaks._

/**
 * @author Ganymede
 */
object GroupTop3 {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setAppName("Top3").setMaster("local[1]")
    val sc = new SparkContext(conf)

    val lines = sc.textFile("D:/scala-eclipse/workspace/spark-study-java/src/main/resources/score.txt", 1)

    val pairs = lines.map { x =>
      {
        val splited = x.split(" ")
        (splited(0), splited(1).toInt)
      }
    }

    val groupedPairs = pairs.groupByKey();

    val top3Score = groupedPairs.map(classScores => {
      val top3 = Array[Int](-1, -1, -1)

      val className = classScores._1

      val scores = classScores._2

      for (score <- scores) {
        breakable {
          for (i <- 0 until 3) {
            if (top3(i) == -1) {
              top3(i) = score;
              break;
            } else if (score > top3(i)) {
              var j = 2
              while (j > i) {
                top3(j) = top3(j - 1);
                j = j - 1
              }
              top3(i) = score;
              break;
            }
          }
        }
      }
      (className, top3);
    })

    top3Score.foreach(x => {
      println(x._1)
      val res = x._2
      for (i <- res) {
        println(i)
      }
      println("==========================")
    })

  }
}

输出:

class1
98
95
90
==========================
class2
96
88
87
========================== 

在实现group by 后的排序算法,用到了 break函数.

scala没有提供类似于java的break语句。但是可以使用boolean类型变量、return或者Breaks的break函数来替代使用。 


2、用spark-sql来实现

创建一个表

create table scores(className string, score int) ROW FORMAT DELIMITED FIELDS TERMINATED BY ' '
加载数据
load data local inpath '/opt/software/tmp/scores.data'  overwrite into table scores;

查询按班级分组并返回倒序的top3

select className,score from (SELECT className,score, Row_Number() OVER (partition by className ORDER BY score desc ) rank FROM scores ) a where a.rank<=3;
实际就是用了 row_number() over (partition by ... order by ...)的函数。同样hive也是支持的


3、总结:实际生产中,大部分还是用SQL来分析与统计的,明显方便一条SQL搞定了;而代码实现更灵活,便于性能的优化。



你可能感兴趣的:(spark,row_number,分组topN)