核心知识-Flink 增量聚合和全窗口函数你知道多少——第九课 - 超级大汇总——DBC的博客超级大汇总—

一、Flink里面你需要知道的AggregateFunction增量聚合函数知识

窗口 window function ，对窗口内的数据做啥？

定义了要对窗口中收集的数据做的计算操作
增量聚合函数

aggregate(agg函数,WindowFunction(){  })

窗口保存临时数据，每进入一个新数据，会与中间数据累加，生成新的中间数据，再保存到窗口中
常见的增量聚合函数有 reduceFunction、aggregateFunction
min、max、sum 都是简单的聚合操作，不需要自定义规则

AggregateFunction<IN, ACC, OUT>
IN是输入类型，ACC是中间聚合状态类型，OUT是输出类型，是聚合统计当前窗口的数据

代码实战

Flink17AggWindowApp

点击查看完整内容

import net.xdclass.model.VideoOrder;
import net.xdclass.source.VideoOrderSourceV2;
import org.apache.flink.api.common.RuntimeExecutionMode;
import org.apache.flink.api.common.functions.AggregateFunction;
import org.apache.flink.api.java.functions.KeySelector;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.windowing.assigners.TumblingProcessingTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;


public class Flink17AggWindowApp {

    /**
     * source
     * transformation
     * sink
     *
     * @param args
     */
    public static void main(String[] args) throws Exception {

        //构建执行任务环境以及任务的启动的入口, 存储全局相关的参数
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);
        env.setParallelism(1);

        //数据源 source
        DataStream<VideoOrder> ds = env.addSource(new VideoOrderSourceV2());

        KeyedStream<VideoOrder, String> keyByDS = ds.keyBy(new KeySelector<VideoOrder, String>() {
            @Override
            public String getKey(VideoOrder value) throws Exception {
                return value.getTitle();
            }
        });

        SingleOutputStreamOperator<VideoOrder> aggregate = keyByDS.window(TumblingProcessingTimeWindows.of(Time.seconds(5)))
                .aggregate(new AggregateFunction<VideoOrder, VideoOrder, VideoOrder>() {


                    //初始化累加器
                    @Override
                    public VideoOrder createAccumulator() {
                        VideoOrder videoOrder = new VideoOrder();
                        return videoOrder;
                    }

                    //聚合操作
                    @Override
                    public VideoOrder add(VideoOrder value, VideoOrder accumulator) {

                        accumulator.setMoney(value.getMoney() + accumulator.getMoney());

                        if (accumulator.getTitle() == null) {
                            accumulator.setTitle(value.getTitle());
                        }
                        if (accumulator.getCreateTime() == null) {
                            accumulator.setCreateTime(value.getCreateTime());
                        }
                        return accumulator;
                    }

                    //获取结果
                    @Override
                    public VideoOrder getResult(VideoOrder accumulator) {
                        return accumulator;
                    }

                    @Override
                    public VideoOrder merge(VideoOrder a, VideoOrder b) {
                        VideoOrder videoOrder = new VideoOrder();
                        videoOrder.setMoney(a.getMoney() + b.getMoney());
                        videoOrder.setTitle(a.getTitle());
                        return videoOrder;
                    }
                });

        aggregate.print();


        //DataStream需要调用execute,可以取个名称
        env.execute("sliding window job");
    }

}

控制台输出

点击查看完整内容

产生:springc loud，价格:20, 时间:2021-11-18 09:43:37
产生:项目大课，价格:1, 时间:2021-11-18 09:43:38
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:43:39
VideoOrder{tradeNo='null', title='flink', money=45, userId=0, createTime=2021-11-18 09:43:35}
VideoOrder{tradeNo='null', title='面试专题第一季', money=50, userId=0, createTime=2021-11-18 09:43:39}
VideoOrder{tradeNo='null', title='springc loud', money=20, userId=0, createTime=2021-11-18 09:43:37}
VideoOrder{tradeNo='null', title='项目大课', money=2, userId=0, createTime=2021-11-18 09:43:36}
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:43:40
产生:kafka，价格:300, 时间:2021-11-18 09:43:41
产生:flink，价格:45, 时间:2021-11-18 09:43:42
产生:springc loud，价格:20, 时间:2021-11-18 09:43:43
产生:springc loud，价格:20, 时间:2021-11-18 09:43:44
VideoOrder{tradeNo='null', title='面试专题第一季', money=50, userId=0, createTime=2021-11-18 09:43:40}
VideoOrder{tradeNo='null', title='springc loud', money=40, userId=0, createTime=2021-11-18 09:43:43}
VideoOrder{tradeNo='null', title='flink', money=45, userId=0, createTime=2021-11-18 09:43:42}
VideoOrder{tradeNo='null', title='kafka', money=300, userId=0, createTime=2021-11-18 09:43:41}
产生:kafka，价格:300, 时间:2021-11-18 09:43:45
产生:kafka，价格:300, 时间:2021-11-18 09:43:46
产生:spring boot，价格:15, 时间:2021-11-18 09:43:47
产生:java，价格:10, 时间:2021-11-18 09:43:48
产生:spring boot，价格:15, 时间:2021-11-18 09:43:49
VideoOrder{tradeNo='null', title='kafka', money=600, userId=0, createTime=2021-11-18 09:43:45}
VideoOrder{tradeNo='null', title='java', money=10, userId=0, createTime=2021-11-18 09:43:48}
VideoOrder{tradeNo='null', title='spring boot', money=30, userId=0, createTime=2021-11-18 09:43:47}
产生:spring boot，价格:15, 时间:2021-11-18 09:43:50
产生:springc loud，价格:20, 时间:2021-11-18 09:43:51
产生:springc loud，价格:20, 时间:2021-11-18 09:43:52
产生:springc loud，价格:20, 时间:2021-11-18 09:43:53
产生:spring boot，价格:15, 时间:2021-11-18 09:43:54
VideoOrder{tradeNo='null', title='spring boot', money=30, userId=0, createTime=2021-11-18 09:43:50}
VideoOrder{tradeNo='null', title='springc loud', money=60, userId=0, createTime=2021-11-18 09:43:51}
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:43:55
产生:kafka，价格:300, 时间:2021-11-18 09:43:56
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:43:57
产生:spring boot，价格:15, 时间:2021-11-18 09:43:58
产生:spring boot，价格:15, 时间:2021-11-18 09:43:59
VideoOrder{tradeNo='null', title='面试专题第一季', money=100, userId=0, createTime=2021-11-18 09:43:55}
VideoOrder{tradeNo='null', title='spring boot', money=30, userId=0, createTime=2021-11-18 09:43:58}
VideoOrder{tradeNo='null', title='kafka', money=300, userId=0, createTime=2021-11-18 09:43:56}
产生:java，价格:10, 时间:2021-11-18 09:44:00
产生:java，价格:10, 时间:2021-11-18 09:44:01
产生:flink，价格:45, 时间:2021-11-18 09:44:02
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:44:03
产生:spring boot，价格:15, 时间:2021-11-18 09:44:04
VideoOrder{tradeNo='null', title='java', money=20, userId=0, createTime=2021-11-18 09:44:00}
VideoOrder{tradeNo='null', title='spring boot', money=15, userId=0, createTime=2021-11-18 09:44:04}
VideoOrder{tradeNo='null', title='面试专题第一季', money=50, userId=0, createTime=2021-11-18 09:44:03}
VideoOrder{tradeNo='null', title='flink', money=45, userId=0, createTime=2021-11-18 09:44:02}
产生:springc loud，价格:20, 时间:2021-11-18 09:44:05
产生:spring boot，价格:15, 时间:2021-11-18 09:44:06
产生:flink，价格:45, 时间:2021-11-18 09:44:07
产生:项目大课，价格:1, 时间:2021-11-18 09:44:08
产生:kafka，价格:300, 时间:2021-11-18 09:44:09
VideoOrder{tradeNo='null', title='springc loud', money=20, userId=0, createTime=2021-11-18 09:44:05}
VideoOrder{tradeNo='null', title='kafka', money=300, userId=0, createTime=2021-11-18 09:44:09}
VideoOrder{tradeNo='null', title='项目大课', money=1, userId=0, createTime=2021-11-18 09:44:08}
VideoOrder{tradeNo='null', title='flink', money=45, userId=0, createTime=2021-11-18 09:44:07}
VideoOrder{tradeNo='null', title='spring boot', money=15, userId=0, createTime=2021-11-18 09:44:06}
产生:kafka，价格:300, 时间:2021-11-18 09:44:10
产生:java，价格:10, 时间:2021-11-18 09:44:11
产生:java，价格:10, 时间:2021-11-18 09:44:12
产生:spring boot，价格:15, 时间:2021-11-18 09:44:13
产生:spring boot，价格:15, 时间:2021-11-18 09:44:14
VideoOrder{tradeNo='null', title='kafka', money=300, userId=0, createTime=2021-11-18 09:44:10}
VideoOrder{tradeNo='null', title='spring boot', money=30, userId=0, createTime=2021-11-18 09:44:13}
VideoOrder{tradeNo='null', title='java', money=20, userId=0, createTime=2021-11-18 09:44:11}
产生:java，价格:10, 时间:2021-11-18 09:44:15
产生:java，价格:10, 时间:2021-11-18 09:44:16
产生:spring boot，价格:15, 时间:2021-11-18 09:44:17
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:44:18
产生:项目大课，价格:1, 时间:2021-11-18 09:44:19
VideoOrder{tradeNo='null', title='java', money=20, userId=0, createTime=2021-11-18 09:44:15}
VideoOrder{tradeNo='null', title='项目大课', money=1, userId=0, createTime=2021-11-18 09:44:19}
VideoOrder{tradeNo='null', title='面试专题第一季', money=50, userId=0, createTime=2021-11-18 09:44:18}
VideoOrder{tradeNo='null', title='spring boot', money=15, userId=0, createTime=2021-11-18 09:44:17}
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:44:20
产生:spring boot，价格:15, 时间:2021-11-18 09:44:21
产生:flink，价格:45, 时间:2021-11-18 09:44:22
产生:springc loud，价格:20, 时间:2021-11-18 09:44:23
产生:项目大课，价格:1, 时间:2021-11-18 09:44:24
VideoOrder{tradeNo='null', title='面试专题第一季', money=50, userId=0, createTime=2021-11-18 09:44:20}
VideoOrder{tradeNo='null', title='项目大课', money=1, userId=0, createTime=2021-11-18 09:44:24}
VideoOrder{tradeNo='null', title='springc loud', money=20, userId=0, createTime=2021-11-18 09:44:23}
VideoOrder{tradeNo='null', title='flink', money=45, userId=0, createTime=2021-11-18 09:44:22}
VideoOrder{tradeNo='null', title='spring boot', money=15, userId=0, createTime=2021-11-18 09:44:21}
产生:面试专题第一季，价格:50, 时间:2021-11-18 09:44:25

Process finished with exit code -1

温馨提示

你可以理解为就是一个自定义的操作，里面你想干什么都可以！

二、Flink里面你需要知道的WindowFunction全窗口函数知识

全窗口函数

apply(new WindowFunction(){ })

窗口先缓存该窗口所有元素，等窗口的全部数据收集起来后再触发条件计算
常见的全窗口聚合函数 windowFunction(未来可能弃用)、processWindowFunction(可以获取到窗口上下文更多信息,包括窗口信息)

IN是输入类型，OUT是输出类型，KEY是分组类型，W是时间窗 
WindowFunction<IN, OUT, KEY, W extends Window>

代码实战

Flink17ApplyWindowApp

点击查看完整内容

import net.xdclass.model.VideoOrder;
import net.xdclass.source.VideoOrderSourceV2;
import org.apache.commons.collections.IteratorUtils;
import org.apache.flink.api.common.RuntimeExecutionMode;
import org.apache.flink.api.common.functions.AggregateFunction;
import org.apache.flink.api.java.functions.KeySelector;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.windowing.WindowFunction;
import org.apache.flink.streaming.api.windowing.assigners.TumblingProcessingTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;
import org.apache.flink.streaming.api.windowing.windows.TimeWindow;
import org.apache.flink.util.Collector;

import java.util.List;
import java.util.stream.Collectors;


public class Flink17ApplyWindowApp {

    /**
     * source
     * transformation
     * sink
     *
     * @param args
     */
    public static void main(String[] args) throws Exception {

        //构建执行任务环境以及任务的启动的入口, 存储全局相关的参数
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);
        env.setParallelism(1);

        //数据源 source
        DataStream<VideoOrder> ds = env.addSource(new VideoOrderSourceV2());

        KeyedStream<VideoOrder, String> keyByDS = ds.keyBy(new KeySelector<VideoOrder, String>() {
            @Override
            public String getKey(VideoOrder value) throws Exception {
                return value.getTitle();
            }
        });


        SingleOutputStreamOperator<VideoOrder> apply = keyByDS.window(TumblingProcessingTimeWindows.of(Time.seconds(5)))
                .apply(new WindowFunction<VideoOrder, VideoOrder, String, TimeWindow>() {
                    @Override
                    public void apply(String key, TimeWindow window, Iterable<VideoOrder> input, Collector<VideoOrder> out) throws Exception {

                        List<VideoOrder> list = IteratorUtils.toList(input.iterator());
                        int total = list.stream().collect(Collectors.summingInt(VideoOrder::getMoney)).intValue();
                        VideoOrder videoOrder = new VideoOrder();
                        videoOrder.setMoney(total);
                        videoOrder.setTitle(list.get(0).getTitle());
                        videoOrder.setCreateTime(list.get(0).getCreateTime());
                        out.collect(videoOrder);

                    }
                });

        apply.print();


        //DataStream需要调用execute,可以取个名称
        env.execute("sliding window job");
    }

}

控制台输出

点击查看完整内容

三、Flink里面你需要知道的processWindowFunction全窗口函数知识

全窗口函数

process(new ProcessWindowFunction(){})

窗口先缓存该窗口所有元素，等窗口的全部数据收集起来后再触发条件计算
常见的全窗口聚合函数 windowFunction(未来可能弃用)、processWindowFunction(可以获取到窗口上下文更多信息，包括窗口信息)

IN是输入类型，OUT是输出类型，KEY是分组类型，W是时间窗 
ProcessWindowFunction<IN, OUT, KEY, W extends Window>

代码实战

Flink17ProcessWindowApp

点击查看完整内容

<pre class="prettyprint lang-java linenums:1">
import net.xdclass.model.VideoOrder;
import net.xdclass.source.VideoOrderSourceV2;
import org.apache.commons.collections.IteratorUtils;
import org.apache.flink.api.common.RuntimeExecutionMode;
import org.apache.flink.api.java.functions.KeySelector;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.windowing.ProcessWindowFunction;
import org.apache.flink.streaming.api.functions.windowing.WindowFunction;
import org.apache.flink.streaming.api.windowing.assigners.TumblingProcessingTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;
import org.apache.flink.streaming.api.windowing.windows.TimeWindow;
import org.apache.flink.util.Collector;

import java.util.List;
import java.util.stream.Collectors;


public class Flink17ProcessWindowApp {

    /**
     * source
     * transformation
     * sink
     *
     * @param args
     */
    public static void main(String[] args) throws Exception {

        //构建执行任务环境以及任务的启动的入口, 存储全局相关的参数
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);
        env.setParallelism(1);

        //数据源 source
        DataStream&lt;VideoOrder&gt; ds = env.addSource(new VideoOrderSourceV2());

        KeyedStream&lt;VideoOrder, String&gt; keyByDS = ds.keyBy(new KeySelector&lt;VideoOrder, String&gt;() {
            @Override
            public String getKey(VideoOrder value) throws Exception {
                return value.getTitle();
            }
        });


        SingleOutputStreamOperator&lt;VideoOrder&gt; process = keyByDS
                .window(TumblingProcessingTimeWindows.of(Time.seconds(5)))
                .process(new ProcessWindowFunction&lt;VideoOrder, VideoOrder, String, TimeWindow&gt;() {
                    @Override
                    public void process(String key, Context context, Iterable&lt;VideoOrder&gt; elements, Collector&lt;VideoOrder&gt; out) throws Exception {

                        List&lt;VideoOrder&gt; list = IteratorUtils.toList(elements.iterator());
                        int total = list.stream().collect(Collectors.summingInt(VideoOrder::getMoney)).intValue();
                        VideoOrder videoOrder = new VideoOrder();
                        videoOrder.setMoney(total);
                        videoOrder.setCreateTime(list.get(0).getCreateTime());
                        videoOrder.setTitle(list.get(0).getTitle());

                        
                        out.collect(videoOrder);

                    }
                });

                process.print();



        //DataStream需要调用execute,可以取个名称
        env.execute(&quot;sliding window job&quot;);
    }

}
</pre>

控制台输出

点击查看完整内容

窗口函数对比

增量聚合

aggregate(new AggregateFunction(){});

全窗口聚合

apply(new WindowFunction(){})

process(new ProcessWindowFunction(){}) //比上面这个强

本文作者为DBC，转载请注明。

Flink