Flink Window——ReduceFunction、AggregateFunction、ProcessWindowFunction窗口函数详解

1.使用 ReduceFunction函数

让两个元素结合起来,产生一个相同类型的元素,它是增量的,放在KeyBy函数之后

package flink.java.test;

import org.apache.flink.api.common.functions.ReduceFunction;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

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

        DataStreamSource<Tuple3<String,String,Integer>> input = env.fromElements(ENGLISH);
        input.keyBy(x -> x.f0)
                .countWindow(2)
                .reduce(new ReduceFunction<Tuple3<String, String, Integer>>() {
                    @Override
                    public Tuple3<String, String, Integer> reduce(Tuple3<String, String, Integer> value1, Tuple3<String, String, Integer> value2) throws Exception {

                        System.out.println("value1-->"+value1);
                        System.out.println("value2-->"+value2);

                        System.out.println("==========================");
                        return new Tuple3<>(value1.f0,value1.f1,value1.f2+value2.f2);
                    }
                }).print("reduce累加");

        env.execute() ;

    }

    public static final Tuple3[] ENGLISH = new Tuple3[]{
            //班级 姓名 成绩
            Tuple3.of("class1","张三",100),
            Tuple3.of("class1","李四",30),
            Tuple3.of("class1","王五",70),
            Tuple3.of("class2","赵六",50),
            Tuple3.of("class2","小七",40),
            Tuple3.of("class2","小八",10),
    };

}

  

执行返回结果:

value1-->(class2,赵六,50)
value2-->(class2,小七,40)
==========================
reduce累加:1> (class2,赵六,90)
value1-->(class1,张三,100)
value2-->(class1,李四,30)
==========================
reduce累加:2> (class1,张三,130)

Process finished with exit code 0

  

2.使用AggregateFunction函数统计计算

AggregateFunction 比 ReduceFunction 更加的通用,它有三个参数,一个输入类型(IN),一个累加器(ACC),一个输出类型(OUT)。

输入类型,就是输入流的类型。接口中有一个方法,可以把输入的元素和累加器累加。并且可以初始化一个累加器,然后把两个累加器合并成一个累加器,获得输出结果。

        input.keyBy(x -> x.f0)
                .countWindow(2)

//        AggregateFunction 比 ReduceFunction 更加的通用,它有三个参数,一个输入类型(IN),一个累加器(ACC),一个输出类型(OUT)
                .aggregate(new AggregateFunction<Tuple3<String, String, Integer>, Tuple2<String,Integer>, Tuple2<String,Integer>>() {
//    创建累加器操作:初始化中间值 @Override
public Tuple2<String, Integer> createAccumulator() { return Tuple2.of("class1",1000); }
//    累加器操作 @Override
public Tuple2<String, Integer> add(Tuple3<String, String, Integer> value1, Tuple2<String, Integer> value2) { return Tuple2.of(value1.f0,value1.f2+value2.f1); }
//    获取结果 @Override
public Tuple2<String, Integer> getResult(Tuple2<String, Integer> value) { return Tuple2.of(value.f0,value.f1); } //    累加器合并操作,只有会话窗口的时候才会调用! @Override public Tuple2<String, Integer> merge(Tuple2<String, Integer> value, Tuple2<String, Integer> acc1) { return Tuple2.of(value.f0,value.f1+acc1.f1); } }) .print("aggregate累加") ; env.execute() ;

执行结果:

aggregate累加:1> (class2,1090)
aggregate累加:2> (class1,1130)

Process finished with exit code 0

  

3.ProcessWindowFunction(全窗口函数)

ProcessWindowFunction 有一个 Iterable 迭代器,用来获得窗口中所有的元素。

有一个上下文对象用来获得时间和状态信息,比其他的窗口函数有更大的灵活性。

但是这样做损耗了一部分性能和资源,因为元素不能增量聚合,相反 ,在触发窗口计算时,Flink 需要在内部缓存窗口的所有元素。

案例1:

        input.keyBy(x -> x.f0)
                .countWindow(2)
                //public abstract class ProcessWindowFunction<IN, OUT, KEY, W extends Window> extends AbstractRichFunction ...
                .process(new ProcessWindowFunction<Tuple3<String, String, Integer>, Tuple3<String,String,Integer>, String, GlobalWindow>() {
                    @Override
                    public void process(String s,   //参数1:key
                                        Context context,    //参数2:上下文对象
                                        Iterable<Tuple3<String, String, Integer>> iterable, //参数3:这个窗口的所有元素
                                        //参数4:收集器,用于向下游传递数据
                                        Collector<Tuple3<String, String, Integer>> collector) throws Exception {
                        System.out.println(context.window().maxTimestamp());
                        int sum = 0 ;
                        String name = "" ;
                        for (Tuple3<String,String,Integer> tuple3:iterable){
                            sum += tuple3.f2 ;
                            name = tuple3.f1 ;
                        }

                        collector.collect(Tuple3.of(s,name,sum));
                    }
                }).print();

  

输出结果:

9223372036854775807
2> (class1,李四,130)
9223372036854775807
1> (class2,小七,90)

Process finished with exit code 0

  

案例2:

.process(new ProcessWindowFunction<Tuple2<String, Long>, Tuple2<String, Long>, String, TimeWindow>() {
    // 参数1: key 参数2: 上下文对象 参数3: 这个窗口内所有的元素 参数4: 收集器, 用于向下游传递数据
    @Override
    public void process(String key,
                        Context context,
                        Iterable<Tuple2<String, Long>> elements,
                        Collector<Tuple2<String, Long>> out) throws Exception {
        System.out.println(context.window().getStart());
        long sum = 0L;
        for (Tuple2<String, Long> t : elements) {
            sum += t.f1;
        }
        out.collect(Tuple2.of(key, sum));
    }
})

参考:

Flink Window那些事——ReduceFunction窗口函数

Flink(14) 窗口函数(window function) 详解

原文地址:https://www.cnblogs.com/-courage/p/14674309.html