3 Spark入门distinct、union、intersection，subtract，cartesian等数学运算

这一篇是一些简单的Spark操作，如去重、合并、取交集等，不管用不用的上，做个档案记录。

distinct去重

import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;

import java.util.Arrays;
import java.util.List;

/**
 * 去除重复的元素，不过此方法涉及到混洗，操作开销很大
 * @author wuweifeng wrote on 2018/4/16.
 */
public class TestDistinct {
    public static void main(String[] args) {
        SparkSession sparkSession = SparkSession.builder().appName("JavaWordCount").master("local").getOrCreate();
        //spark对普通List的reduce操作
        JavaSparkContext javaSparkContext = new JavaSparkContext(sparkSession.sparkContext());
        List<Integer> data = Arrays.asList(1, 1, 2, 3, 4, 5);
        JavaRDD<Integer> originRDD = javaSparkContext.parallelize(data);
        List<Integer> results = originRDD.distinct().collect();
        System.out.println(results);
    }
}

结果是[4, 1, 3, 5, 2]

union合并，不去重

这个就是简单的将两个RDD合并到一起

import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;

import java.util.Arrays;
import java.util.List;

/**
 * 合并两个RDD
 * @author wuweifeng wrote on 2018/4/16.
 */
public class TestUnion {
    public static void main(String[] args) {
        SparkSession sparkSession = SparkSession.builder().appName("JavaWordCount").master("local").getOrCreate();
        //spark对普通List的reduce操作
        JavaSparkContext javaSparkContext = new JavaSparkContext(sparkSession.sparkContext());
        List<Integer> one = Arrays.asList(1, 2, 3, 4, 5);
        List<Integer> two = Arrays.asList(1, 6, 7, 8, 9);
        JavaRDD<Integer> oneRDD = javaSparkContext.parallelize(one);
        JavaRDD<Integer> twoRDD = javaSparkContext.parallelize(two);
        List<Integer> results = oneRDD.union(twoRDD).collect();
        System.out.println(results);
    }
}

结果是[1, 2, 3, 4, 5, 1, 6, 7, 8, 9]

intersection取交集

import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;

import java.util.Arrays;
import java.util.List;

/**
 * 返回两个RDD的交集
 * @author wuweifeng wrote on 2018/4/16.
 */
public class TestIntersection {
    public static void main(String[] args) {
        SparkSession sparkSession = SparkSession.builder().appName("JavaWordCount").master("local").getOrCreate();
        //spark对普通List的reduce操作
        JavaSparkContext javaSparkContext = new JavaSparkContext(sparkSession.sparkContext());
        List<Integer> one = Arrays.asList(1, 2, 3, 4, 5);
        List<Integer> two = Arrays.asList(1, 6, 7, 8, 9);
        JavaRDD<Integer> oneRDD = javaSparkContext.parallelize(one);
        JavaRDD<Integer> twoRDD = javaSparkContext.parallelize(two);
        List<Integer> results = oneRDD.intersection(twoRDD).collect();
        System.out.println(results);
    }
}

结果[1]

subtract

RDD1.subtract(RDD2)，返回在RDD1中出现，但是不在RDD2中出现的元素，不去重 

import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;

import java.util.Arrays;
import java.util.List;

/**
 * @author wuweifeng wrote on 2018/4/16.
 */
public class TestSubtract {
    public static void main(String[] args) {
        SparkSession sparkSession = SparkSession.builder().appName("JavaWordCount").master("local").getOrCreate();
        //spark对普通List的reduce操作
        JavaSparkContext javaSparkContext = new JavaSparkContext(sparkSession.sparkContext());
        List<Integer> one = Arrays.asList(1, 2, 3, 4, 5);
        List<Integer> two = Arrays.asList(1, 6, 7, 8, 9);
        JavaRDD<Integer> oneRDD = javaSparkContext.parallelize(one);
        JavaRDD<Integer> twoRDD = javaSparkContext.parallelize(two);

        List<Integer> results = oneRDD.subtract(twoRDD).collect();
        System.out.println(results);
    }
}

结果：[2, 3, 4, 5]

cartesian返回笛卡尔积

笛卡尔积就是两两组合的所有组合，这个的开销非常大，譬如A是["a","b","c"]，B是["1","2","3"]，那笛卡尔积就是

( 1 a)( 1 b)( 1 c)( 2 a)( 2 b)( 2 c)( 3 a)( 3 b)( 3 c)

import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.sql.SparkSession;
import scala.Tuple2;

import java.util.Arrays;
import java.util.List;

/**
 * 返回笛卡尔积，开销很大
 * @author wuweifeng wrote on 2018/4/16.
 */
public class TestCartesian {
    public static void main(String[] args) {
        SparkSession sparkSession = SparkSession.builder().appName("JavaWordCount").master("local").getOrCreate();
        //spark对普通List的reduce操作
        JavaSparkContext javaSparkContext = new JavaSparkContext(sparkSession.sparkContext());
        List<Integer> one = Arrays.asList(1, 2, 3);
        List<Integer> two = Arrays.asList(1, 4, 5);
        JavaRDD<Integer> oneRDD = javaSparkContext.parallelize(one);
        JavaRDD<Integer> twoRDD = javaSparkContext.parallelize(two);
        List<Tuple2<Integer, Integer>> results = oneRDD.cartesian(twoRDD).collect();
        System.out.println(results);
    }
}

注意，返回的是键值对

[(1,1), (1,4), (1,5), (2,1), (2,4), (2,5), (3,1), (3,4), (3,5)]