1. 案例

用户ID,活动ID,时间,事件类型,省份

u001,A1,2019-09-02 10:10:11,1,北京市

u001,A1,2019-09-02 14:10:11,1,北京市

u001,A1,2019-09-02 14:10:11,2,北京市

u002,A1,2019-09-02 14:10:11,1,北京市

u002,A2,2019-09-02 14:10:11,1,北京市

u002,A2,2019-09-02 15:10:11,1,北京市

u002,A2,2019-09-02 15:10:11,2,北京市

事件类型：

  0：曝光

  1：点击

  2：参与

需求：统计点击、参与某个活动的人数和次数
  • 方案一：使用ValueState结合HashSet实现

　具体代码如下

ActivityCountAdv1

package cn._51doit.flink.day08;

import org.apache.flink.api.common.functions.MapFunction;

import org.apache.flink.api.common.state.ValueState;

import org.apache.flink.api.common.state.ValueStateDescriptor;

import org.apache.flink.api.common.typeinfo.TypeHint;

import org.apache.flink.api.common.typeinfo.TypeInformation;

import org.apache.flink.api.java.tuple.Tuple;

import org.apache.flink.api.java.tuple.Tuple4;

import org.apache.flink.api.java.tuple.Tuple5;

import org.apache.flink.configuration.Configuration;

import org.apache.flink.streaming.api.datastream.DataStreamSource;

import org.apache.flink.streaming.api.datastream.KeyedStream;

import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;

import org.apache.flink.streaming.api.environment.LocalStreamEnvironment;

import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

import org.apache.flink.streaming.api.functions.KeyedProcessFunction;

import org.apache.flink.util.Collector;

import java.util.HashSet;

public class ActivityCountAdv1 {

    public static void main(String[] args) throws Exception {

        LocalStreamEnvironment env = StreamExecutionEnvironment.createLocalEnvironment();

        DataStreamSource<String> lines = env.socketTextStream("feng05", 8888);

        // 对数据进行切分整理

        SingleOutputStreamOperator<Tuple5<String, String, String, Integer, String>> tpDataStream = lines.map(new MapFunction<String, Tuple5<String, String, String, Integer, String>>() {

            @Override

            public Tuple5<String, String, String, Integer, String> map(String line) throws Exception {

                String[] fields = line.split(",");

                String uid = fields[0];

                String activityID = fields[1];

                String date = fields[2];

                Integer type = Integer.parseInt(fields[3]);

                String prince = fields[4];

                return Tuple5.of(uid, activityID, date, type, prince);

            }

        });

        // 按照活动ID和事件类型分组

        KeyedStream<Tuple5<String, String, String, Integer, String>, Tuple> keyed = tpDataStream.keyBy(1, 3);

        keyed.process(new KeyedProcessFunction<Tuple, Tuple5<String, String, String, Integer, String>, Tuple4<String, Integer, Integer, Integer>>() {

            //保存去重后用户ID的HashSet

            private transient ValueState<HashSet<String>> uidState;

            //保存次数的Integer类型

            private transient ValueState<Integer> countState;

            @Override

            public void open(Configuration parameters) throws Exception {

                // 定义一个状态描述器

                ValueStateDescriptor<HashSet<String>> stateDescriptor1 = new ValueStateDescriptor<HashSet<String>>(

                        "uid-state",

                        TypeInformation.of(new TypeHint<HashSet<String>>(){})

                );

                // 定义一个状态描述器

                ValueStateDescriptor<Integer> stateDescriptor2 = new ValueStateDescriptor<Integer>(

                        "count-state",

                        Integer.class

                );

                // 获取状态

                uidState = getRuntimeContext().getState(stateDescriptor1);

                countState = getRuntimeContext().getState(stateDescriptor2);

            }

            @Override

            public void processElement(Tuple5<String, String, String, Integer, String> value, Context ctx, Collector<Tuple4<String, Integer, Integer, Integer>> out) throws Exception {

                String uid = value.f0;

                String aid = value.f1;

                Integer type = value.f3;

                //使用HashSet进行判断去重,更新uidState

                HashSet<String> hashSet = uidState.value();

                if(hashSet == null){

                    hashSet = new HashSet<>();

                }

                hashSet.add(uid);

                uidState.update(hashSet);

                // 计算人数

                Integer count = countState.value();

                if(count == null) {

                    count = 0;

                }

                count += 1;

                countState.update(count);

                out.collect(Tuple4.of(aid,type,hashSet.size(), count));

            }

        }).print();

        env.execute();

    }

}

　　如果使用HashSet去重，用户实例较大，会大量消耗资源，导致性能变低，甚至内存溢出

  • 方案二：改进，使用BloomFilter存储用户的ID，BloomFilter可以判断用户一定不存在，使用的内存极少。但是使用BloomFilter没有计数器，就必须额外定义一个状态，存储去重的人数

ActivityCountAdv2

package cn._51doit.flink.day08;

import org.apache.flink.api.common.functions.MapFunction;

import org.apache.flink.api.common.state.ValueState;

import org.apache.flink.api.common.state.ValueStateDescriptor;

import org.apache.flink.api.common.typeinfo.TypeHint;

import org.apache.flink.api.common.typeinfo.TypeInformation;

import org.apache.flink.api.java.tuple.Tuple;

import org.apache.flink.api.java.tuple.Tuple4;

import org.apache.flink.api.java.tuple.Tuple5;

import org.apache.flink.configuration.Configuration;

import org.apache.flink.shaded.guava18.com.google.common.hash.BloomFilter;

import org.apache.flink.shaded.guava18.com.google.common.hash.Funnels;

import org.apache.flink.streaming.api.datastream.DataStreamSource;

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.KeyedProcessFunction;

import org.apache.flink.util.Collector;

import java.nio.charset.Charset;

import java.nio.charset.StandardCharsets;

import java.util.HashSet;

public class ActivityCountAdv2 {

    public static void main(String[] args) throws Exception {

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        //u001,A1,2019-09-02 10:10:11,1,北京市

        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        //对数据进行切分整理

        SingleOutputStreamOperator<Tuple5<String, String, String, String, String>> tpDataStream = lines.map(new MapFunction<String, Tuple5<String, String, String, String, String>>() {

            @Override

            public Tuple5<String, String, String, String, String> map(String line) throws Exception {

                String[] fields = line.split(",");

                String uid = fields[0];

                String aid = fields[1];

                String time = fields[2];

                String type = fields[3];

                String province = fields[4];

                return Tuple5.of(uid, aid, time, type, province);

            }

        });

        //按照活动ID和事件类型分组

        KeyedStream<Tuple5<String, String, String, String, String>, Tuple> keyed = tpDataStream.keyBy(1, 3);

        keyed.process(new KeyedProcessFunction<Tuple, Tuple5<String, String, String, String, String>, Tuple4<String, String, Integer, Integer>>() {

            //保存去重后用户ID的HashSet

            private transient ValueState<BloomFilter> uidState;

            //保存用户ID去重的次数的Integer类型

            private transient ValueState<Integer> uidCountState;

            //保存次数的Integer类型（未去重的）

            private transient ValueState<Integer> countState;

            @Override

            public void open(Configuration parameters) throws Exception {

                //定义一个状态描述器

                ValueStateDescriptor<BloomFilter> stateDescriptor1 = new ValueStateDescriptor<BloomFilter>(

                        "uid-state",

                        TypeInformation.of(new TypeHint<BloomFilter>(){})

                );

                //定义一个状态描述器

                ValueStateDescriptor<Integer> stateDescriptor2 = new ValueStateDescriptor<Integer>(

                        "count-state",

                        Integer.class

                );

                //定义一个状态描述器

                ValueStateDescriptor<Integer> stateDescriptor3 = new ValueStateDescriptor<Integer>(

                        "uid-count-state",

                        Integer.class

                );

                //获取状态

                //获取状态

                uidState = getRuntimeContext().getState(stateDescriptor1);

                countState = getRuntimeContext().getState(stateDescriptor2);

                uidCountState = getRuntimeContext().getState(stateDescriptor3);

            }

            @Override

            public void processElement(Tuple5<String, String, String, String, String> value, Context ctx, Collector<Tuple4<String, String, Integer, Integer>> out) throws Exception {

                String uid = value.f0;

                String aid = value.f1;

                String type = value.f3;

                //使用HashSet进行判断去重

                BloomFilter bloomFilter = uidState.value();

                Integer uidCount = uidCountState.value(); //人数

                Integer count = countState.value(); //次数

                if(count == null) {

                    count = 0;

                }

                if(bloomFilter == null) {

                    bloomFilter = BloomFilter.create(Funnels.unencodedCharsFunnel(), 10000000);

                    uidCount = 0;

                }

                if(!bloomFilter.mightContain(uid)) {

                    bloomFilter.put(uid); //添加到BloomFilter中

                    uidCount += 1;

                }

                count += 1;

                countState.update(count);

                uidState.update(bloomFilter);

                uidCountState.update(uidCount);

                out.collect(Tuple4.of(aid, type, uidCount, count));

            }

        }).print();

        env.execute();

    }

}

 2. 活动指标多维度统计

　　此处要进行多次key操作（一中维度就需要keyBy一次），相当繁琐。此处是通过将数据存入redis，所以不需要使用flink中的state，具体见代码

ActivityCountWithMultiDimension

package cn._51doit.flink.day08;

import org.apache.flink.api.common.functions.MapFunction;

import org.apache.flink.api.java.tuple.Tuple3;

import org.apache.flink.api.java.utils.ParameterTool;

import org.apache.flink.streaming.api.datastream.DataStreamSource;

import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;

import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

import sun.awt.geom.AreaOp;

public class ActivityCountWithMultiDimension {

    public static void main(String[] args) throws Exception{

        ParameterTool parameters = ParameterTool.fromPropertiesFile(args[0]);

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        env.getConfig().setGlobalJobParameters(parameters);

        //u001,A1,2019-09-02 10:10:11,1,北京市

        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        SingleOutputStreamOperator<ActivityBean> beanStream = lines.map(new MapFunction<String, ActivityBean>() {

            @Override

            public ActivityBean map(String line) throws Exception {

                String[] fields = line.split(",");

                String uid = fields[0];

                String aid = fields[1];

                String date = fields[2].split(" ")[0];

                String type = fields[3];

                String province = fields[4];

                return ActivityBean.of(uid, aid, date, type, province);

            }

        });

        SingleOutputStreamOperator<ActivityBean> res1 = beanStream.keyBy("aid", "type").sum("count");

        SingleOutputStreamOperator<ActivityBean> res2 = beanStream.keyBy("aid", "type", "date").sum("count");

        SingleOutputStreamOperator<ActivityBean> res3 = beanStream.keyBy("aid", "type", "date", "province").sum("count");

        res1.map(new MapFunction<ActivityBean, Tuple3<String, String, String>>() {

            @Override

            public Tuple3<String, String, String> map(ActivityBean value) throws Exception {

                return Tuple3.of(Constant.ACTIVITY_COUNT +"-"+ value.aid,  value.type, value.count.toString());

            }

        }).addSink(new MyRedisSink());

        res2.map(new MapFunction<ActivityBean, Tuple3<String, String, String>>() {

            @Override

            public Tuple3<String, String, String> map(ActivityBean value) throws Exception {

                return Tuple3.of(Constant.DAILY_ACTIVITY_COUNT + "-" + value.aid + "-" + value.date, value.type, value.count.toString());

            }

        }).addSink(new MyRedisSink());

        res3.map(new MapFunction<ActivityBean, Tuple3<String, String, String>>() {

            @Override

            public Tuple3<String, String, String> map(ActivityBean value) throws Exception {

                return Tuple3.of(Constant.PROVINCE_DAILY_ACTIVITY_COUNT + "-" + value.aid + "-" + value.date + "-" + value.province, value.type, value.count.toString());

            }

        }).addSink(new MyRedisSink());

        env.execute();

    }

}

Constant

package cn._51doit.flink.day08;

public class Constant {

    public static final String ACTIVITY_COUNT = "ACTIVITY_COUNT";

    public static final String DAILY_ACTIVITY_COUNT = "DAILY_ACTIVITY_COUNT";

    public static final String PROVINCE_DAILY_ACTIVITY_COUNT = "PROVINCE_DAILY_ACTIVITY_COUNT";

}

ActivityBean

package cn._51doit.flink.day08;

public class ActivityBean {

    public String uid;

    public String aid;

    public String date;

    public String type;

    public String province;

    public Long count = 1L;

    public ActivityBean() {}

    public ActivityBean(String uid, String aid, String date, String type, String province) {

        this.uid = uid;

        this.aid = aid;

        this.date = date;

        this.type = type;

        this.province = province;

    }

    public static ActivityBean of(String uid, String aid, String date, String type, String province) {

        return new ActivityBean(uid, aid, date, type, province);

    }

    @Override

    public String toString() {

        return "ActivityBean{" +

                "uid='" + uid + '\'' +

                ", aid='" + aid + '\'' +

                ", date='" + date + '\'' +

                ", type='" + type + '\'' +

                ", province='" + province + '\'' +

                '}';

    }

}

MyRedisSink

package cn._51doit.flink.day08;

import org.apache.flink.api.java.tuple.Tuple3;

import org.apache.flink.api.java.utils.ParameterTool;

import org.apache.flink.configuration.Configuration;

import org.apache.flink.streaming.api.functions.sink.RichSinkFunction;

import redis.clients.jedis.Jedis;

public class MyRedisSink extends RichSinkFunction<Tuple3<String, String, String>> {

    private transient Jedis jedis;

    @Override

    public void open(Configuration parameters) throws Exception {

       ParameterTool params = (ParameterTool) getRuntimeContext()

               .getExecutionConfig()

               .getGlobalJobParameters();

        String host = params.getRequired("redis.host");

        String password = params.getRequired("redis.password");

        int port = params.getInt("redis.port", 6379);

        int db = params.getInt("redis.db", 0);

        Jedis jedis = new Jedis(host, port);

        jedis.auth(password);

        jedis.select(db);

        this.jedis = jedis;

    }

    @Override

    public void invoke(Tuple3<String, String, String> value, Context context) throws Exception {

        if (!jedis.isConnected()) {

            jedis.connect();

        }

        jedis.hset(value.f0, value.f1, value.f2);

    }

    @Override

    public void close() throws Exception {

        jedis.close();

    }

}

flink-----实时项目---day04-------1. 案例:统计点击、参与某个活动的人数和次数 2. 活动指标多维度统计（自定义redisSink）的更多相关文章

  1. 5.Flink实时项目之业务数据准备

    1. 流程介绍 在上一篇文章中,我们已经把客户端的页面日志,启动日志,曝光日志分别发送到kafka对应的主题中.在本文中,我们将把业务数据也发送到对应的kafka主题中. 通过maxwell采集业务数 ...

  2. 9.Flink实时项目之订单宽表

    1.需求分析 订单是统计分析的重要的对象,围绕订单有很多的维度统计需求,比如用户.地区.商品.品类.品牌等等.为了之后统计计算更加方便,减少大表之间的关联,所以在实时计算过程中将围绕订单的相关数据整合 ...

  3. 7.Flink实时项目之独立访客开发

    1.架构说明 在上6节当中,我们已经完成了从ods层到dwd层的转换,包括日志数据和业务数据,下面我们开始做dwm层的任务. DWM 层主要服务 DWS,因为部分需求直接从 DWD 层到DWS 层中间 ...

  4. 10.Flink实时项目之订单维度表关联

    1. 维度查询 在上一篇中,我们已经把订单和订单明细表join完,本文将关联订单的其他维度数据,维度关联实际上就是在流中查询存储在 hbase 中的数据表.但是即使通过主键的方式查询,hbase 速度 ...

  5. 11.Flink实时项目之支付宽表

    支付宽表 支付宽表的目的,最主要的原因是支付表没有到订单明细,支付金额没有细分到商品上, 没有办法统计商品级的支付状况. 所以本次宽表的核心就是要把支付表的信息与订单明细关联上. 解决方案有两个 一个 ...

  6. 3.Flink实时项目之流程分析及环境搭建

    1. 流程分析 前面已经将日志数据(ods_base_log)及业务数据(ods_base_db_m)发送到kafka,作为ods层,接下来要做的就是通过flink消费kafka 的ods数据,进行简 ...

  7. 4.Flink实时项目之数据拆分

    1. 摘要 我们前面采集的日志数据已经保存到 Kafka 中,作为日志数据的 ODS 层,从 kafka 的ODS 层读取的日志数据分为 3 类, 页面日志.启动日志和曝光日志.这三类数据虽然都是用户 ...

  8. 6.Flink实时项目之业务数据分流

    在上一篇文章中,我们已经获取到了业务数据的输出流,分别是dim层维度数据的输出流,及dwd层事实数据的输出流,接下来我们要做的就是把这些输出流分别再流向对应的数据介质中,dim层流向hbase中,dw ...

  9. 1.Flink实时项目前期准备

    1.日志生成项目 日志生成机器:hadoop101 jar包:mock-log-0.0.1-SNAPSHOT.jar gmall_mock ​ |----mock_common ​ |----mock ...

随机推荐

  1. Manacher算法 求 最长回文子串

    1 概述(扯淡) 在了解Manacher算法之前,我们得先知道什么是回文串和子串. 回文串,就是正着看反着看都一样的字符串.比如说"abba"就是一个回文串,"abbc& ...

  2. i love dingning

    "如果你爱一个人,不是下课给人家买买水,不是短信发来发去,也不是周末一起出来唱唱歌聊聊天吃吃饭,而是做一个出色的人.以后的以后,可能还有别的人爱她,你要做的,是把别人都比下去.你要变得优秀, ...

  3. IDEA升级开源框架

    在开发过程中,我们经常会用到一些 GitHub或者Gitee上的开源框架来快速搭建我们的业务系统,但是当框架被我们大批量修改后,开源框架又有升级了.这时候升级框架就变得很麻烦,也不能直接直接进行合并, ...

  4. Django（74）drf-spectacular自动生成接口文档

    介绍 drf-spectacular是为Django REST Framework生成合理灵活的OpenAPI 3.0模式.它可以自动帮我们提取接口中的信息,从而形成接口文档,而且内容十分详细,再也不 ...

  5. DeWeb --- Hello,World!

    1.新建一个DLL,命名为hello.dpr 2.新增一个Form.(File->New->VCL Form - Delphi),建议不要更改单元名称和Form名称,即分别为unit1.p ...

  6. shell 脚本控制命令的执行顺序

    &&,||,(),{},& 五个符号的运用shell脚本执行命令的时候,有时候会依赖于前一个命令是否执行成功.而&&和||就是用来判断前一个命令执行效果的. 也 ...

  7. Typecho部署小破站

    写在前面 以前利用 Github Page + Hexo框架 + Next主题搭建过静态博客,没错就是那个黑白色系的网页!但是体验并不是很好,一来本身是静态网页,页面内容要修改都需要在本地修改完上传到 ...

  8. 组件通过props属性传值

    组件之间的传值 组件是一个单独功能模块的封装,有属于自己的data和methods,一个组件的 data 选项必须是一个函数 为什么必须是函数:因为只有当data是函数时,不同实例调用同一个组件时才会 ...

  9. JVM 核心参数

    JVM 内存相关的几个核心参数 参数部分看我笔记   https://note.youdao.com/s/Ch3awnVu JVM模板 1. ParNew + CMS 版 根据服务调整 -Xmx -X ...

  10. 手把手教你学Dapr - 7. Actors

    上一篇:手把手教你学Dapr - 6. 发布订阅 介绍 Actor模式将Actor描述为最低级别的"计算单元".换句话说,您在一个独立的单元(称为actor)中编写代码,该单元接收 ...