Flink DataSet partitionByRange sortPartition 用法 實例

package DataSetPartitionTest1;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.operators.Order;
import org.apache.flink.api.java.DataSet;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * @Author you guess
 * @Date 2020/7/4 11:58
 * @Version 1.0
 * @Desc
 */
public class DataSetPartitionTest1 {

    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        DataSet<Tuple2<Long, String>> source2 = env.fromElements(
                Tuple2.of(2L, "xiaoli"),
                Tuple2.of(1L, "shinelon"),
                Tuple2.of(67L, "hhhhhh67"),
                Tuple2.of(2L, "xiaohong"),
                Tuple2.of(3L, "hhhhhh"),
                Tuple2.of(13L, "hhhhhh13"),
                Tuple2.of(51L, "hhhhhh51"),
                Tuple2.of(89L, "hhhhhh89"),
                Tuple2.of(23L, "hhhhhh23"),
                Tuple2.of(49L, "hhhhhh49"),
                Tuple2.of(3L, "hhhhhh37"));

        System.out.println("-----111111111111--------");


        //partitionByRange作用是把key相同的元素放入同一個分區（由同一個線程處理）
        //注意：是相同的key，不是hash值相同的key
        source2.partitionByRange(0).sortPartition(0, Order.DESCENDING).map(new MapFunction<Tuple2<Long, String>, Object>() {
            @Override
            public Object map(Tuple2<Long, String> value) throws Exception {
                System.out.println("線程id#" + Thread.currentThread().getId() + "," + value.f0);
                return new Tuple2<Long, String>(value.f0, value.f1);
            }
        }).print();


        System.out.println("-----222222222--------");


        //sortPartition默認由一個線程處理，是全局有效，保證整個數據集有序
        source2.sortPartition(0, Order.DESCENDING).map(new MapFunction<Tuple2<Long, String>, Object>() {
            @Override
            public Object map(Tuple2<Long, String> value) throws Exception {
                System.out.println("線程id#" + Thread.currentThread().getId());
                return new Tuple2<Long, String>(value.f0, value.f1);
            }
        }).print();

        System.out.println("---3333333333---");

        //先按第一個字段排升序，再按第二個字段排升序，= order by c1,c2 ;默認asc;
        source2.sortPartition(0, Order.ASCENDING).sortPartition(1, Order.ASCENDING).print();

        System.out.println("---4444444---");
        //設置並行度是2，則是每個並行度內有序，並不是整個數據集有序
        source2.sortPartition(0, Order.DESCENDING).setParallelism(2).map(new MapFunction<Tuple2<Long, String>, Object>() {
            @Override
            public Object map(Tuple2<Long, String> value) throws Exception {
                System.out.println("線程id#" + Thread.currentThread().getId() + "," + value.f0);
                return new Tuple2<Long, String>(value.f0, value.f1);
            }
        }).print();

    }
}


/*
-----111111111111--------
線程id#97,2
線程id#98,3
線程id#100,13
線程id#98,3
線程id#97,2
線程id#101,23
線程id#102,49
線程id#103,51
線程id#104,67
線程id#105,89
線程id#95,1
(1,shinelon)
(2,xiaoli)
(2,xiaohong)
(3,hhhhhh)
(3,hhhhhh37)
(13,hhhhhh13)
(23,hhhhhh23)
(49,hhhhhh49)
(51,hhhhhh51)
(67,hhhhhh67)
(89,hhhhhh89)
-----222222222--------
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
線程id#238
(89,hhhhhh89)
(67,hhhhhh67)
(51,hhhhhh51)
(49,hhhhhh49)
(23,hhhhhh23)
(13,hhhhhh13)
(3,hhhhhh)
(3,hhhhhh37)
(2,xiaoli)
(2,xiaohong)
(1,shinelon)
---3333333333---
(1,shinelon)
(2,xiaohong)
(2,xiaoli)
(3,hhhhhh)
(3,hhhhhh37)
(13,hhhhhh13)
(23,hhhhhh23)
(49,hhhhhh49)
(51,hhhhhh51)
(67,hhhhhh67)
(89,hhhhhh89)
---4444444---
線程id#403,67
線程id#406,89
線程id#403,51
線程id#403,23
線程id#403,3
線程id#403,3
線程id#403,2
線程id#406,49
線程id#406,13
線程id#406,2
線程id#406,1
(67,hhhhhh67)
(51,hhhhhh51)
(23,hhhhhh23)
(3,hhhhhh)
(3,hhhhhh37)
(2,xiaoli)
(89,hhhhhh89)
(49,hhhhhh49)
(13,hhhhhh13)
(2,xiaohong)
(1,shinelon)
*/

flink 1.9.2,java1.8 

sortPartition的默認並行度是1，是全局有序，
或者 
sortPartition(0, Order.DESCENDING).setParallelism(1)並行度是1，也是全局有序，

Flink DataSet first groupBy sortGroup 用法 實例