11、TopN实战

1、Java版本：

    1.1、取前3

package sparkcore.java;

import java.util.List;

import org.apache.spark.SparkConf;

import org.apache.spark.api.java.JavaPairRDD;

import org.apache.spark.api.java.JavaRDD;

import org.apache.spark.api.java.JavaSparkContext;

import org.apache.spark.api.java.function.Function;

import org.apache.spark.api.java.function.PairFunction;

import scala.Tuple2;

/**

 * 取前3

 */

public class Top3 {

    public static void main(String[] args) {

        SparkConf conf = new SparkConf()

                .setAppName("Top3")

                .setMaster("local");  

        JavaSparkContext sc = new JavaSparkContext(conf);

    /*

         3

        5

        6

        7

        1

        4

        5

        6

        9

        0

        3

     */

        JavaRDD<String> lines = sc.textFile("top.txt");

        

        JavaPairRDD<Integer, String> pairs = lines.mapToPair(

                

                new PairFunction<String, Integer, String>() {

                    private static final long serialVersionUID = 1L;

                    @Override

                    public Tuple2<Integer, String> call(String t) throws Exception {

                        return new Tuple2<Integer, String>(Integer.valueOf(t), t);

                    }

                    

                });

        

        JavaPairRDD<Integer, String> sortedPairs = pairs.sortByKey(false);

        

        JavaRDD<Integer> sortedNumbers = sortedPairs.map(

                

                new Function<Tuple2<Integer,String>, Integer>() {

                    private static final long serialVersionUID = 1L;

                    @Override

                    public Integer call(Tuple2<Integer, String> v1) throws Exception {

                        return v1._1;

                    }

                    

                });

        

        List<Integer> sortedNumberList = sortedNumbers.take(3);

        

        for(Integer num : sortedNumberList) {

            System.out.println(num);

        }

        

        sc.close();

    }

}

输出：

9

7

6

    1.2、每组内取top3

package sparkcore.java;

import java.util.ArrayList;

import java.util.Arrays;

import java.util.Collections;

import java.util.Comparator;

import java.util.Iterator;

import java.util.List;

import org.apache.spark.SparkConf;

import org.apache.spark.api.java.JavaPairRDD;

import org.apache.spark.api.java.JavaRDD;

import org.apache.spark.api.java.JavaSparkContext;

import org.apache.spark.api.java.function.PairFunction;

import org.apache.spark.api.java.function.VoidFunction;

import scala.Tuple2;

/**

 * 每组内取top3

 */

public class GroupTop3 {

    public static void main(String[] args) {

        SparkConf conf = new SparkConf().setAppName("Top3").setMaster("local");

        JavaSparkContext sc = new JavaSparkContext(conf);

        /*

             class1 90

            class2 56

            class1 87

            class1 76

            class2 88

            class1 95

            class1 74

            class2 87

            class2 67

            class2 77

         */

        JavaRDD<String> lines = sc.textFile("score.txt");

        

        JavaPairRDD<String, Integer> pairs = lines.mapToPair(

                new PairFunction<String, String, Integer>() {

                    private static final long serialVersionUID = 1L;

                    @Override

                    public Tuple2<String, Integer> call(String line) throws Exception {

                        String[] lineSplited = line.split(" ");

                        return new Tuple2<String, Integer>(lineSplited[0], Integer.valueOf(lineSplited[1]));

                    }

                });

        JavaPairRDD<String, Iterable<Integer>> groupedPairs = pairs.groupByKey();

        

        JavaPairRDD<String, Iterable<Integer>> top3Score = groupedPairs.mapToPair(

                new PairFunction<Tuple2<String, Iterable<Integer>>, String, Iterable<Integer>>() {

                    private static final long serialVersionUID = 1L;

                    @Override

                    public Tuple2<String, Iterable<Integer>> call(Tuple2<String, Iterable<Integer>> classScores)

                            throws Exception {

                        Integer[] top3 = new Integer[3];

                        String className = classScores._1;

                        Iterator<Integer> scores = classScores._2.iterator();

                        List<Integer> l = new ArrayList<Integer>();

                        while (scores.hasNext()) {

                            l.add(scores.next());

                        }

                        Collections.sort(l, new Comparator<Integer>() {

                            public int compare(Integer o1, Integer o2) {

                                return -o1.compareTo(o2);

                            }

                        });

                        for (int i = 0; i < l.size(); i++) {

                            if (i < 3) {

                                top3[i] = l.get(i);

                            }

                        }

                        return new Tuple2<String, Iterable<Integer>>(className, Arrays.asList(top3));

                    }

                });

        top3Score.foreach(new VoidFunction<Tuple2<String, Iterable<Integer>>>() {

            private static final long serialVersionUID = 1L;

            @Override

            public void call(Tuple2<String, Iterable<Integer>> t) throws Exception {

                System.out.println("class: " + t._1);

                Iterator<Integer> scoreIterator = t._2.iterator();

                while (scoreIterator.hasNext()) {

                    Integer score = scoreIterator.next();

                    System.out.println(score);

                }

                System.out.println("=======================================");

            }

        });

        sc.close();

    }

}

输出：

class: class1

95

90

87

=======================================

class: class2

88

87

77

=======================================

2、Scala版本：

    2.1、取前3

package sparkcore.scala

import org.apache.spark.SparkConf

import org.apache.spark.SparkContext

object Top3 {

  def main(args: Array[String]): Unit = {

    val conf = new SparkConf()

      .setAppName("Top3")

      .setMaster("local")

    val sc = new SparkContext(conf)

    val lines = sc.textFile("top.txt", 1)

    val pairs = lines.map { line => (line.toInt, line) }

    val sortedPairs = pairs.sortByKey(false)

    val sortedNumbers = sortedPairs.map(sortedPair => sortedPair._1)

    val top3Number = sortedNumbers.take(3)

    for (num <- top3Number) {

      println(num)

    }

  }

}

输出：

9

7

6

    2.2、每组内取top3

package sparkcore.scala

import org.apache.spark.SparkConf

import org.apache.spark.SparkContext

object GroupTop3 {

  def main(args: Array[String]): Unit = {

    val conf = new SparkConf()

      .setAppName("Top3")

      .setMaster("local")

    val sc = new SparkContext(conf)

    val lines = sc.textFile("score.txt", 1)

    val linePair = lines.map((line => (line.split(" ")(0), line.split(" ")(1).toInt)))

    val groupRdd = linePair.groupByKey()

    val sortRdd = groupRdd.map(g => (g._1,

      g._2.toList.sorted(new Ordering[Int]() {

        def compare(x: Int, y: Int): Int = {

          - x.compare(y)

        }

      }).take(3)))

    sortRdd.foreach(e => println(e._1 + " : " + e._2))

  }

}

输出：

class1 : List(95, 90, 87)

class2 : List(88, 87, 77)