flink 1.10.1 cep java版本案例（实时统计每个商户当日订单数和交易金额）

本文的基础环境可以参考flink 1.10.1 java版本wordcount演示 (nc + socket)

这里采用静态数据进行模拟商户订单数据。当接收到订单数据时，会根据商户id进行分组，然后再按天进行统计。

1. 主程序代码

package com.demo.statis;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.streaming.api.TimeCharacteristic;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.AssignerWithPeriodicWatermarks;
import org.apache.flink.streaming.api.watermark.Watermark;

import javax.annotation.Nullable;

public class ShopDayGMVStatis {

    public static void main(String[] args) throws Exception {
        //创建执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        // StreamExecutionEnvironment env = StreamExecutionEnvironment.createLocalEnvironmentWithWebUI(new Configuration());
        env.setParallelism(1);

        // 处理时间
        env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime);

        //从文件中读取数据
        String inputPath = "data/shoporder.txt";

        DataStream<String> inputDataSet = env.readTextFile(inputPath);

        //对数据集进行处理
        DataStream<ShopOrder> dataStream = inputDataSet.map(new MapFunction<String, ShopOrder>() {
            @Override
            public ShopOrder map(String s) throws Exception {
                String splits[] = s.split(",");
                Long time = new Long(splits[1]);
                return new ShopOrder(splits[0], time * 1000,
                        new Long(splits[2]) );
            }
        })
                .assignTimestampsAndWatermarks(
                        new AssignerWithPeriodicWatermarks<ShopOrder>()
                        {
                            @Override
                            public long extractTimestamp(ShopOrder shopOrder, long l) {
                                return shopOrder.getPayTime();
                            }

                            @Nullable
                            @Override
                            public Watermark getCurrentWatermark() {
                                return null;
                            }
                        }
                );

        DataStream<ShopDayGmv> result = dataStream.keyBy(ShopOrder::getShopId)
                .process(new ShopDayKeyProcess());

        result.print("shop day statistics");

        env.execute();
    }
}

这里根据订单支付时间的对应日期进行统计。 

2. 统计代码

package com.demo.statis;

import org.apache.flink.api.common.state.MapState;
import org.apache.flink.api.common.state.MapStateDescriptor;
import org.apache.flink.api.common.state.StateTtlConfig;
import org.apache.flink.api.common.time.Time;
import org.apache.flink.api.common.typeinfo.TypeInformation;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.KeyedProcessFunction;
import org.apache.flink.util.Collector;

import java.text.SimpleDateFormat;
import java.util.Date;


public class ShopDayKeyProcess extends KeyedProcessFunction<String, ShopOrder, ShopDayGmv> {

    // MapState key:date+type, value:amount。例如：key="18581Sales" 标示2020-11-15 的销售额
    private MapState<String, Long> dayAmount;

    @Override
    public void open(Configuration parameters) throws Exception {
        super.open(parameters);

        // ttl策略，(此处需要注意，有些参数跟随flink设置而略有不同)
        StateTtlConfig stateTtlConfig = StateTtlConfig.newBuilder(Time.days(3))
                .setStateVisibility(StateTtlConfig.StateVisibility.NeverReturnExpired)
                .setUpdateType(StateTtlConfig.UpdateType.OnCreateAndWrite)
                .cleanupInRocksdbCompactFilter(3000)
                .build();
        MapStateDescriptor<String, Long> amountStateDesc = new MapStateDescriptor<>("amountState", TypeInformation.of(String.class), TypeInformation.of(Long.class));

        // 设置ttl
        amountStateDesc.enableTimeToLive(stateTtlConfig);

        // 获取MapState
        dayAmount = getRuntimeContext().getMapState(amountStateDesc);

    }

    @Override
    public void processElement(ShopOrder value, Context context, Collector<ShopDayGmv> collector) throws Exception {
        final ShopDayGmv shopDayGmv = new ShopDayGmv();

        shopDayGmv.setShopId(value.getShopId());
        shopDayGmv.setTime(value.getPayTime());

        long payTime = value.getPayTime();
        // paytime，下面公式计算得到日期标示，此值不会重复
        // 60 * 60 * 24 = 86400(秒) = 1天
        // 60 * 60 * 8 = 28800(秒) = 8小时，对应到GMT+8小时
        long dateNum = (payTime + 28800) / 86400;

        shopDayGmv.setTimeKey(timeFormat(payTime));

        // 从状态中获取每日销售额
        Long dateSalesState = dayAmount.get(dateNum + "Sales");
        Long dateCountState = dayAmount.get(dateNum + "Count");

        // 销售额
        if (dateSalesState == null) {
            // 没有数据，标示该时期下第一笔订单，订单量=1，销售额=price
            dayAmount.put(dateNum + "Sales", value.getOrderPrice());
        } else {
            dayAmount.put(dateNum + "Sales", value.getOrderPrice() + dateSalesState);
        }
        // 订单量
        if (dateCountState == null) {
            // 没有数据，标示该时期下第一笔订单，订单量=1，销售额=price
            dayAmount.put(dateNum + "Count", 1L);
        } else {
            dayAmount.put(dateNum + "Count", 1L + dateCountState);
        }
        // 获取每日订单量和销售额
        shopDayGmv.setTodayCount(dayAmount.get(dateNum + "Count"));
        shopDayGmv.setTodaySales(dayAmount.get(dateNum + "Sales"));

        // 数据收集
        collector.collect(shopDayGmv);
    }

    public static String timeFormat(long timeStamp) {
        return new SimpleDateFormat("yyyy-MM-dd").format(new Date(timeStamp));
    }

}

通过设置状态缓存，每个商户的统计状态数据只会保存3天，这样可以避免数据的不断累积导致内存被消耗完的问题。 

3. 辅助代码ShopOrder

package com.demo.statis;


public class ShopOrder {

    private String shopId;
    private Long payTime;
    private Long orderPrice;

    public ShopOrder() {
    }

    public ShopOrder(String shopId, Long payTime, Long orderPrice) {
        this.shopId = shopId;
        this.payTime = payTime;
        this.orderPrice = orderPrice;
    }

    public String getShopId() {
        return shopId;
    }

    public void setShopId(String shopId) {
        this.shopId = shopId;
    }

    public Long getPayTime() {
        return payTime;
    }

    public void setPayTime(Long payTime) {
        this.payTime = payTime;
    }

    public Long getOrderPrice() {
        return orderPrice;
    }

    public void setOrderPrice(Long orderPrice) {
        this.orderPrice = orderPrice;
    }

    @Override
    public String toString() {
        return "ShopOrder{" +
                "shopId='" + shopId + '\'' +
                ", payTime=" + payTime +
                ", orderPrice=" + orderPrice +
                '}';
    }
}

4. 辅助代码ShopDayGmv

package com.demo.statis;


public class ShopDayGmv {

    private String shopId;
    private Long time;
    private Long todayCount;
    private Long todaySales;
    private String timeKey;

    public ShopDayGmv() {
    }

    public ShopDayGmv(String shopId, Long time, Long todayCount, Long todaySales, String timeKey) {
        this.shopId = shopId;
        this.time = time;
        this.todayCount = todayCount;
        this.todaySales = todaySales;
        this.timeKey = timeKey;
    }

    public String getShopId() {
        return shopId;
    }

    public void setShopId(String shopId) {
        this.shopId = shopId;
    }

    public Long getTime() {
        return time;
    }

    public void setTime(Long time) {
        this.time = time;
    }

    public Long getTodayCount() {
        return todayCount;
    }

    public void setTodayCount(Long todayCount) {
        this.todayCount = todayCount;
    }

    public Long getTodaySales() {
        return todaySales;
    }

    public void setTodaySales(Long todaySales) {
        this.todaySales = todaySales;
    }

    public String getTimeKey() {
        return timeKey;
    }

    public void setTimeKey(String timeKey) {
        this.timeKey = timeKey;
    }

    @Override
    public String toString() {
        return "ShopDayGmv{" +
                "shopId='" + shopId + '\'' +
                ", time=" + time +
                ", todayCount=" + todayCount +
                ", todaySales=" + todaySales +
                ", timeKey='" + timeKey + '\'' +
                '}';
    }
}

5. 测试数据

001,1644479447,10
001,1644479447,20
001,1644479445,10
001,1644479446,30
001,1644479447,20
001,1644479444,80
001,1644479445,10
002,1641801046,10
003,1641801056,30
004,1641801066,10

6. 执行程序

可以看到如下的输出结果：

shop day statistics> ShopDayGmv{shopId='001', time=1644479447000, todayCount=1, todaySales=10, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479447000, todayCount=2, todaySales=30, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479445000, todayCount=3, todaySales=40, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479446000, todayCount=4, todaySales=70, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479447000, todayCount=5, todaySales=90, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479444000, todayCount=6, todaySales=170, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='001', time=1644479445000, todayCount=7, todaySales=180, timeKey='2022-02-10'}
shop day statistics> ShopDayGmv{shopId='002', time=1641801046000, todayCount=1, todaySales=10, timeKey='2022-01-10'}
shop day statistics> ShopDayGmv{shopId='003', time=1641801056000, todayCount=1, todaySales=30, timeKey='2022-01-10'}
shop day statistics> ShopDayGmv{shopId='004', time=1641801066000, todayCount=1, todaySales=10, timeKey='2022-01-10'}

可以看出，已经安装商户编号和日志进行了统计汇总输出。这里测试只是输出到控制台，实际项目中可以根据项目情况将结果输出到kafka、redis或者数据库等。

可以看到，针对每个数据，都会出现一个对应的输出结果进行对应。如果输入数据并发比较大的话，输出的数据并发也会相应增大，此时需要选择合适的输出方式。