groupByKey,reduceByKey,sortByKey算子-Java&Python版Spark

Posted 2020-08-27

groupByKey,reduceByKey,sortByKey算子

 

视频教程：

1、优酷

2、 YouTube

 

1、groupByKey

 groupByKey是对每个key进行合并操作，但只生成一个sequence，groupByKey本身不能自定义操作函数。

java：

 1 package com.bean.spark.trans;
 2 
 3 import java.util.Arrays;
 4 import java.util.List;
 5 
 6 import org.apache.spark.SparkConf;
 7 import org.apache.spark.api.java.JavaPairRDD;
 8 import org.apache.spark.api.java.JavaSparkContext;
 9 
10 import scala.Tuple2;
11 
12 public class TraGroupByKey {
13     public static void main(String[] args) {
14         SparkConf conf = new SparkConf();
15         conf.setMaster("local");
16         conf.setAppName("union");
17         System.setProperty("hadoop.home.dir", "D:/tools/spark-2.0.0-bin-hadoop2.6");
18         JavaSparkContext sc = new JavaSparkContext(conf);
19         List<Tuple2<String, Integer>> list = Arrays.asList(new Tuple2<String, Integer>("cl1", 90),
20                 new Tuple2<String, Integer>("cl2", 91),new Tuple2<String, Integer>("cl3", 97),
21                 new Tuple2<String, Integer>("cl1", 96),new Tuple2<String, Integer>("cl1", 89),
22                 new Tuple2<String, Integer>("cl3", 90),new Tuple2<String, Integer>("cl2", 60));
23         JavaPairRDD<String, Integer> listRDD = sc.parallelizePairs(list);
24         JavaPairRDD<String, Iterable<Integer>> results  = listRDD.groupByKey();
25         System.out.println(results.collect());
26         sc.close();
27     }
28 }

python：

 1 # -*- coding:utf-8 -*-
 2 
 3 from pyspark import SparkConf
 4 from pyspark import SparkContext
 5 import os
 6 
 7 if __name__ == ‘__main__‘:
 8     os.environ["SPARK_HOME"] = "D:/tools/spark-2.0.0-bin-hadoop2.6"
 9     conf = SparkConf().setMaster(‘local‘).setAppName(‘group‘)
10     sc = SparkContext(conf=conf)
11     data = [(‘tom‘,90),(‘jerry‘,97),(‘luck‘,92),(‘tom‘,78),(‘luck‘,64),(‘jerry‘,50)]
12     rdd = sc.parallelize(data)
13     print rdd.groupByKey().map(lambda x: (x[0],list(x[1]))).collect()

注意：当采用groupByKey时，由于它不接收函数，spark只能先将所有的键值对都移动，这样的后果是集群节点之间的开销很大，导致传输延时。

整个过程如下：

因此，在对大数据进行复杂计算时，reduceByKey优于groupByKey。

另外，如果仅仅是group处理，那么以下函数应该优先于 groupByKey ：

（1）、combineByKey 组合数据，但是组合之后的数据类型与输入时值的类型不一样。

（2）、foldByKey合并每一个 key 的所有值，在级联函数和“零值”中使用。

2、reduceByKey

对数据集key相同的值，都被使用指定的reduce函数聚合到一起。

java：

 

 1 package com.bean.spark.trans;
 2 
 3 import java.util.Arrays;
 4 import java.util.List;
 5 
 6 import org.apache.spark.SparkConf;
 7 import org.apache.spark.api.java.JavaPairRDD;
 8 import org.apache.spark.api.java.JavaSparkContext;
 9 import org.apache.spark.api.java.function.Function2;
10 
11 import scala.Tuple2;
12 
13 public class TraReduceByKey {
14     public static void main(String[] args) {
15         SparkConf conf = new SparkConf();
16         conf.setMaster("local");
17         conf.setAppName("reduce");
18         System.setProperty("hadoop.home.dir", "D:/tools/spark-2.0.0-bin-hadoop2.6");
19         JavaSparkContext sc = new JavaSparkContext(conf);
20         List<Tuple2<String, Integer>> list = Arrays.asList(new Tuple2<String, Integer>("cl1", 90),
21                 new Tuple2<String, Integer>("cl2", 91),new Tuple2<String, Integer>("cl3", 97),
22                 new Tuple2<String, Integer>("cl1", 96),new Tuple2<String, Integer>("cl1", 89),
23                 new Tuple2<String, Integer>("cl3", 90),new Tuple2<String, Integer>("cl2", 60));
24         JavaPairRDD<String, Integer> listRDD = sc.parallelizePairs(list);
25         JavaPairRDD<String, Integer> results = listRDD.reduceByKey(new Function2<Integer, Integer, Integer>() {
26             @Override
27             public Integer call(Integer s1, Integer s2) throws Exception {
28                 // TODO Auto-generated method stub
29                 return s1 + s2;
30             }
31         });
32         System.out.println(results.collect());
33 sc.close();    
34 }
35 }

python：

 1 # -*- coding:utf-8 -*-
 2 
 3 from pyspark import SparkConf
 4 from pyspark import SparkContext
 5 import os
 6 from operator import add
 7 if __name__ == ‘__main__‘:
 8     os.environ["SPARK_HOME"] = "D:/tools/spark-2.0.0-bin-hadoop2.6"
 9     conf = SparkConf().setMaster(‘local‘).setAppName(‘reduce‘)
10     sc = SparkContext(conf=conf)
11     data = [(‘tom‘,90),(‘jerry‘,97),(‘luck‘,92),(‘tom‘,78),(‘luck‘,64),(‘jerry‘,50)]
12     rdd = sc.parallelize(data)        
13 print rdd.reduceByKey(add).collect()
14 sc.close()

当采用reduceByKey时，Spark可以在每个分区移动数据之前将待输出数据与一个共用的key结合。 注意在数据对被搬移前同一机器上同样的key是怎样被组合的。

 

3、sortByKey

通过key进行排序。

java:

 1 package com.bean.spark.trans;
 2 
 3 import java.util.Arrays;
 4 import java.util.List;
 5 
 6 import org.apache.spark.SparkConf;
 7 import org.apache.spark.api.java.JavaPairRDD;
 8 import org.apache.spark.api.java.JavaSparkContext;
 9 
10 import scala.Tuple2;
11 
12 public class TraSortByKey {
13     public static void main(String[] args) {
14         SparkConf conf = new SparkConf();
15         conf.setMaster("local");
16         conf.setAppName("sort");
17         System.setProperty("hadoop.home.dir", "D:/tools/spark-2.0.0-bin-hadoop2.6");
18         JavaSparkContext sc = new JavaSparkContext(conf);
19         List<Tuple2<Integer, String>> list = Arrays.asList(new Tuple2<Integer,String>(3,"Tom"),
20                 new Tuple2<Integer,String>(2,"Jerry"),new Tuple2<Integer,String>(5,"Luck")
21                 ,new Tuple2<Integer,String>(1,"Spark"),new Tuple2<Integer,String>(4,"Storm"));
22         JavaPairRDD<Integer,String> rdd = sc.parallelizePairs(list);
23         JavaPairRDD<Integer, String> results = rdd.sortByKey(false);
24         System.out.println(results.collect());
25       sc.close()  
26     }
27 }    

python:

1 #-*- coding:utf-8 -*-
2 if __name__ == ‘__main__‘:
3     os.environ["SPARK_HOME"] = "D:/tools/spark-2.0.0-bin-hadoop2.6"
4     conf = SparkConf().setMaster(‘local‘).setAppName(‘reduce‘)
5     sc = SparkContext(conf=conf)
6     data = [(5,90),(1,92),(3,50)]
7     rdd = sc.parallelize(data) 
8 print rdd.sortByKey(False).collect()
9 sc.close()