Flink系列-6、Flink DataSet的Transformation

版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。

大数据系列文章目录

官方网址：https://flink.apache.org/

学习资料：https://flink-learning.org.cn/

Flink 算子

dataSet包括一系列的Transformation操作：
https://ci.apache.org/projects/flink/flink-docs-release-1.10/dev/batch/dataset_transformations.html

Map

将DataSet中的每一个元素转换为另外一个元素

示例
使用map操作，读取apache.log文件中的字符串数据转换成ApacheLogEvent对象
如：

86.149.9.216 10001 17/05/2015:10:05:30 GET /presentations/logstash-monitorama-2013/images/github-contributions.png
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:06:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:07:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:08:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:09:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:10:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:16:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css

步骤

• 获取ExecutionEnvironment运行环境
• 使用readTextFile读取数据构建数据源
• 创建一个ApacheLogEvent类
• 使用map操作执行转换
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.MapOperator;

import java.text.SimpleDateFormat;

/**
 * 演示Flink map算子
 */
public class MapDemo {
    public static void main(String[] args) throws Exception {
        // 0. Env
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        // 1. Source
        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // 2. 使用map 转换字符串为 JavaBean对象
        MapOperator<String, ApacheEventLog> result = logSource.map(new MapFunction<String, ApacheEventLog>() {
            // 注意: 当前的日期转换对象, 是运行在TaskManager中的 也就是被分布式执行的
            // 构建日期转换 17/05/2015:10:05:30
            final SimpleDateFormat inputSDF = new SimpleDateFormat("dd/MM/yyyy:HH:mm:ss");
            final SimpleDateFormat outSDF = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");

            @Override
            public ApacheEventLog map(String value) throws Exception {
                String[] arr = value.split(" ");
                String ip = arr[0];
                String userID = arr[1];
                String date = outSDF.format(inputSDF.parse(arr[2]).getTime());
                String method = arr[3];
                String url = arr[4];
                ApacheEventLog apacheEventLog = new ApacheEventLog(ip, userID, date, method, url);
                return apacheEventLog;
            }
        });

        result.print();
    }

    public static class ApacheEventLog{
        private String ip;
        private String userID;
        private String date;
        private String method;
        private String url;

        public ApacheEventLog() {}

        public ApacheEventLog(String ip, String userID, String date, String method, String url) {
            this.ip = ip;
            this.userID = userID;
            this.date = date;
            this.method = method;
            this.url = url;
        }

        public String getIp() {
            return ip;
        }

        public void setIp(String ip) {
            this.ip = ip;
        }

        public String getUserID() {
            return userID;
        }

        public void setUserID(String userID) {
            this.userID = userID;
        }

        public String getDate() {
            return date;
        }

        public void setDate(String date) {
            this.date = date;
        }

        public String getMethod() {
            return method;
        }

        public void setMethod(String method) {
            this.method = method;
        }

        public String getUrl() {
            return url;
        }

        public void setUrl(String url) {
            this.url = url;
        }

        @Override
        public String toString() {
            return "ApacheEventLog{" +
                    "ip='" + ip + '\'' +
                    ", userID='" + userID + '\'' +
                    ", date='" + date + '\'' +
                    ", method='" + method + '\'' +
                    ", url='" + url + '\'' +
                    '}';
        }
    }
}

FlatMap

将DataSet中的每一个元素转换为0…n个元素

示例

读取flatmap.log文件中的数据

如：

张三,苹果手机,联想电脑,华为平板
李四,华为手机,苹果电脑,小米平板

转换为

张三有苹果手机
张三有联想电脑
张三有华为平板
李四有…
…
…

思路
以上数据为一条转换为三条，显然，应当使用flatMap来实现分别在flatMap函数中构建三个数据，并放入到一个列表中

步骤

• 构建批处理运行环境
• 构建本地集合数据源
• 使用flatMap将一条数据经过处理转换为三条数据
• 使用逗号分隔字段
• 分别构建三条数据
• 打印输出

package batch.transformation;

import org.apache.flink.api.common.functions.FlatMapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.FlatMapOperator;
import org.apache.flink.util.Collector;

/**
 * 演示FlatMap
 * 需求是将一行数据变成多行返回的时候不要嵌套list
 * 可以用flatMap
 */
public class FlatMapDemo {
    public static void main(String[] args) throws Exception {
        // 0. Env
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        // env.setParallelism(1);

        // 1. Source
        DataSource<String> fileSource = env.readTextFile("data/input/flatmap.log");

        // 2. flatMap转换
        FlatMapOperator<String, String> result = fileSource.flatMap(new FlatMapFunction<String, String>() {
            @Override
            /**
             * flatMap 没有返回值
             * 多行的输出就通过调用Collector对象的collect方法进行传递即可
             */
            public void flatMap(String value, Collector<String> out) throws Exception {
                String[] arr = value.split(",");
                String name = arr[0];

                out.collect(name + "有" + arr[1]);
                out.collect(name + "有" + arr[2]);
                out.collect(name + "有" + arr[3]);
            }
        });

        result.print();
    }
}

Filter

过滤出来一些符合条件的元素

示例
读取apache.log文件中的访问日志数据，过滤出来以下访问IP是83.149.9.216的访问日志。

86.149.9.216 10001 17/05/2015:10:05:30 GET /presentations/logstash-monitorama-2013/images/github-contributions.png
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:06:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:07:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:08:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:09:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:10:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:16:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css

步骤

• 获取ExecutionEnvironment运行环境
• 使用fromCollection构建数据源
• 使用filter操作执行过滤
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.FilterFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.FilterOperator;

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

        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // filter过滤
        FilterOperator<String> result = logSource.filter(new FilterFunction<String>() {
            @Override
            public boolean filter(String value) throws Exception {
                String ip = value.split(" ")[0];

                return ip.equals("83.149.9.216");
            }
        });

        result.print();
    }
}

Reduce

可以对一个 dataset 或者一个 group 来进行聚合计算，最终聚合成一个元素

示例

读取apache.log日志，统计ip地址访问pv数量，使用 reduce 操作聚合成一个最终结果

86.149.9.216 10001 17/05/2015:10:05:30 GET /presentations/logstash-monitorama-2013/images/github-contributions.png
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:06:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:07:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:08:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:09:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:10:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:16:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css

结果类似：
(86.149.9.216,1)
(10.0.0.1,7)
(83.149.9.216,6)

步骤

• 获取 ExecutionEnvironment 运行环境
• 使用 readTextFile 构建数据源
• 使用 reduce 执行聚合操作
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.functions.ReduceFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.operators.ReduceOperator;
import org.apache.flink.api.java.operators.UnsortedGrouping;
import org.apache.flink.api.java.tuple.Tuple2;

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

        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // 提取IP, 后面都跟上1(作为元组返回)
        MapOperator<String, Tuple2<String, Integer>> ipWithOne = logSource.map(new MapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String value) throws Exception {
                String ip = value.split(" ")[0];
                return Tuple2.of(ip, 1);
            }
        });

        // 分组 + reduce聚合
        UnsortedGrouping<Tuple2<String, Integer>> grouped = ipWithOne.groupBy(0);

        // reduce 聚合
        ReduceOperator<Tuple2<String, Integer>> result = grouped.reduce(new ReduceFunction<Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> reduce(Tuple2<String, Integer> value1, Tuple2<String, Integer> value2) throws Exception {
                return Tuple2.of(value1.f0, value1.f1 + value2.f1);
            }
        });

        result.print();
    }
}

reduceGroup

可以对一个dataset或者一个group来进行聚合计算，最终聚合成一个元素
reduce和reduceGroup的 区别

首先groupBy函数会将一个个的单词进行分组，分组后的数据被reduce一个个的拉取过来，这种方式如果数据量大的情况下，拉取的数据会非常多，增加了网络IO

reduceGroup是reduce的一种优化方案；
它会先分组reduce，然后在做整体的reduce；这样做的好处就是可以减少网络IO；

示例

读取apache.log日志，统计ip地址访问pv数量，使用 reduceGroup 操作聚合成一个最终结果

步骤

• 获取 ExecutionEnvironment 运行环境
• 使用 readTextFile 构建数据源
• 使用 groupBy 按照单词进行分组
• 使用 reduceGroup 对每个分组进行统计
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.GroupReduceFunction;
import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.functions.ReduceFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.*;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.util.Collector;

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

        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // 提取IP, 后面都跟上1(作为元组返回)
        MapOperator<String, Tuple2<String, Integer>> ipWithOne = logSource.map(new MapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String value) throws Exception {
                String ip = value.split(" ")[0];
                return Tuple2.of(ip, 1);
            }
        });

        // 分组
        UnsortedGrouping<Tuple2<String, Integer>> grouped = ipWithOne.groupBy(0);

        // reduceGroup聚合: 一次聚合一整个分区 节省海量的网络IO请求次数
        GroupReduceOperator<Tuple2<String, Integer>, Tuple2<String, Integer>> result = grouped.reduceGroup(new GroupReduceFunction<Tuple2<String, Integer>, Tuple2<String, Integer>>() {
            @Override
            public void reduce(Iterable<Tuple2<String, Integer>> allGroupData, Collector<Tuple2<String, Integer>> out) throws Exception {
                String key = "";
                int counter = 0;

                for (Tuple2<String, Integer> tuple : allGroupData) {
                    key = tuple.f0;

                    counter += tuple.f1;
                }

                out.collect(Tuple2.of(key, counter));
            }
        });

        result.print();
    }
}

Aggregate

按照内置的方式来进行聚合。例如：SUM/MIN/MAX…

示例

读取apache.log日志，统计ip地址访问pv数量，使用 aggregate 操作进行PV访问量统计

86.149.9.216 10001 17/05/2015:10:05:30 GET /presentations/logstash-monitorama-2013/images/github-contributions.png
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:06:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:07:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:08:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:09:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:10:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:16:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css

步骤

• 获取 ExecutionEnvironment 运行环境
• 使用 readTextFile 构建数据源
• 使用 groupBy 按照单词进行分组
• 使用 aggregate 对每个分组进行 SUM 统计
• 打印测试

reduceGroupedSource.aggregate(Aggregations.MAX, 1);

Aggregate只能作用于元组上

如图：注意，只可用于元组进行Aggregate

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.aggregation.Aggregations;
import org.apache.flink.api.java.operators.AggregateOperator;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.operators.UnsortedGrouping;
import org.apache.flink.api.java.tuple.Tuple2;

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

        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // 提取IP, 后面都跟上1(作为元组返回)
        MapOperator<String, Tuple2<String, Integer>> ipWithOne = logSource.map(new MapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String value) throws Exception {
                String ip = value.split(" ")[0];
                return Tuple2.of(ip, 1);
            }
        });

        // 分组
        UnsortedGrouping<Tuple2<String, Integer>> grouped = ipWithOne.groupBy(0);

        // 求和操作
        AggregateOperator<Tuple2<String, Integer>> sumResult = grouped.aggregate(Aggregations.SUM, 1);
        // 这个就是aggregate算子的快捷写法
        AggregateOperator<Tuple2<String, Integer>> sumResult2 = grouped.sum(1);

        // 求最大值 最小值
        // print方法中 传入字符串可以作为输出的前缀, minBy maxBy
        sumResult.aggregate(Aggregations.MIN, 1).print("最小值");
        sumResult.aggregate(Aggregations.MAX, 1).print("最大值");
        // 快捷写法
        sumResult2.min(1);
        sumResult2.max(1);


       env.execute();
    }
}

Aggregate的简写形式

注意：aggregate有简写的形式，比如：
reduceGroupedSource.aggregate(Aggregations.MAX, 1);
可以写成reduceGroupedSource.max(1);

max方法本质上还是调用的aggregate方法, 这是一种简单写法
min, sum 同理

从源码中可见, max方法还是调用的aggregate

minBy和maxBy

获取指定字段的最大值、最小值

示例

读取apache.log日志，统计ip地址访问pv数量，使用 minBy、maxBy操作进行PV访问量统计

86.149.9.216 10001 17/05/2015:10:05:30 GET /presentations/logstash-monitorama-2013/images/github-contributions.png
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
83.149.9.216 10002 17/05/2015:10:06:53 GET /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:06:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:07:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:08:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:09:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:10:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:16:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css
10.0.0.1 10003 17/05/2015:10:26:53 POST /presentations/logstash-monitorama-2013/css/print/paper.css

步骤

• 获取 ExecutionEnvironment 运行环境
• 使用 fromCollection 构建数据源
• 使用 groupBy 按照单词进行分组
• 使用 maxBy、minBy对每个分组进行操作
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.functions.ReduceFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.operators.ReduceOperator;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * @author lwh
 * @date 2023/4/12
 * @description
 **/
public class MinByMaxByDemo2 {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        DataSource<String> textFileSource = env.readTextFile("data/input/apache.log");

        MapOperator<String, Tuple2<String, Integer>> ipWithOne = textFileSource.map(new MapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String value) throws Exception {
                return Tuple2.of(value.split(" ")[0], 1);
            }
        });

        ReduceOperator<Tuple2<String, Integer>> reduced = ipWithOne.groupBy(0).reduce(new ReduceFunction<Tuple2<String, Integer>>() {
            @Override
            public Tuple2<String, Integer> reduce(Tuple2<String, Integer> value1, Tuple2<String, Integer> value2) throws Exception {
                return Tuple2.of(value1.f0, value1.f1 + value2.f1);
            }
        });

        reduced.minBy(1).print();
        reduced.maxBy(1).print();

    }
}

Aggregate 和 minBy maxBy的区别

Aggregate的min 和 min 方式 对比minBy 和maxBy的区别在：
以min和minBy举例：

首先： 都只能应用于元组数据

另外最重要的区别在于，计算逻辑不同，尽管都是求最小值，但是：
Min在计算的过程中，会记录最小值，对于其它的列，会取最后一次出现的，然后和最小值组合形成结果返回
minBy在计算的过程中，当遇到最小值后，将第一次出现的最小值所在的整个元素返回。

package batch.transformation;

import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * Yanshi minBy 和 aggregate.min的区别
 */
public class MinByVSAggregateMinDemo {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        DataSource<Tuple2<Integer, Integer>> source = env.fromElements(
                Tuple2.of(3, 2),
                Tuple2.of(1, 2),
                Tuple2.of(2, 3),
                Tuple2.of(111, 1),
                Tuple2.of(1, 1),
                Tuple2.of(3, 1),
                Tuple2.of(0, 1),
                Tuple2.of(33, 2)
        );

        // 聚合求最小
        // aggregate的min计算
        source.min(1).print("agg的min:");
        source.minBy(1).print("minBy:");

        /*
        aggregate的min 和 max: 找到最小值和最大值后, 拼接最后一条数据的其它元素, 组合成结果返回
        minBy或者maxBy: 找到第一条出现的最小值 或者最大值 将这一条数据作为结果返回
        所以, minBy或者maxBy的结果更加准确, 一般我们追求结果集的完整选择它们
        如果只想要最大或者最小值本身, 对结果集的其它内容无所谓, 可以用agg的min和max
         */
        env.execute();
    }
}

distinct去重

去除重复的数据

示例
读取apache.log日志，统计有哪些ip访问了网站

步骤

• 获取 ExecutionEnvironment 运行环境
• 使用 readTextFile 构建数据源
• 使用 distinct 指定按照哪个字段来进行去重
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.DistinctOperator;
import org.apache.flink.api.java.operators.MapOperator;

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

        DataSource<String> logSource = env.readTextFile("data/input/apache.log");

        // 提取全部的ip
        MapOperator<String, String> ips = logSource.map(new MapFunction<String, String>() {
            @Override
            public String map(String value) throws Exception {
                return value.split(" ")[0];
            }
        });

        // 对ip进行去重操作
        DistinctOperator<String> result = ips.distinct();

        result.print();
    }
}
/*
distinct算子可以完成 ```全局```去重. ( 撇除分区的影响, 进行整体去重)
 */

注意：distinct(字段index)只可用于tuple类型

Join

使用join可以将两个DataSet连接起来

示例

有两个csv文件，有一个为 score.csv ，一个为 subject.csv ，分别保存了成绩数据以及学科数据

需要将这两个数据连接到一起，然后打印出来。

步骤

• 分别将两个文件复制到项目中的 data/join/input 中
• 构建批处理环境
• 创建两个类
  学科Subject（学科ID、学科名字）
  成绩Score（唯一ID、学生姓名、学科ID、分数——Double类型）
• 分别使用 readCsvFile 加载csv数据源，并制定泛型
• 使用join连接两个DataSet，并使用 where 、 equalTo 方法设置关联条件
• 打印关联后的数据源

package batch.transformation;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.JoinOperator;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * 演示Flink的Join算子
 * 对两个DataSet进行关联, 形成一个DataSet返回
 */
public class JoinDemo {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        // read 学科数据
        DataSource<String> subjectSource = env.readTextFile("data/input/subject.csv");
        // read score
        DataSource<String> scoreSource = env.readTextFile("data/input/score.csv");

        // 将两份数据集 转变成JavaBean
        MapOperator<String, Subject> subject = subjectSource.map(new MapFunction<String, Subject>() {
            @Override
            public Subject map(String value) throws Exception {
                String[] arr = value.split(",");
                return new Subject(Integer.parseInt(arr[0]), arr[1]);
            }
        });
        MapOperator<String, Score> score = scoreSource.map(new MapFunction<String, Score>() {
            @Override
            public Score map(String value) throws Exception {
                String[] arr = value.split(",");
                return new Score(Integer.parseInt(arr[0]), arr[1], Integer.parseInt(arr[2]), Double.parseDouble(arr[3]));
            }
        });

        // 对两个数据集进行join
        // where表示join左边的数据集, equalTo表示右边的数据集
        // Join完成后形成一个二元元组对象返回, 元组中 第一个列是 左边数据集, 第二个列是右边的数据集
        JoinOperator.DefaultJoin<Score, Subject> joined = score.join(subject).where("subjectID").equalTo("id");

        // 对于左外和右外可以使用 leftOuterJoin和 rightOuterJoin 写法和join一致
        // 我们的join算子是 ```内关联```模式
        score.leftOuterJoin(subject).where("subjectID").equalTo("id");

        // 全外关联也支持 笛卡尔积
        score.fullOuterJoin(subject);

        // 将学生分数中的学科id 替换为学科名称
        MapOperator<Tuple2<Score, Subject>, String> result = joined.map(new MapFunction<Tuple2<Score, Subject>, String>() {
            @Override
            public String map(Tuple2<Score, Subject> value) throws Exception {
                int sid = value.f0.id;
                String sname = value.f0.name;
                String subjectName = value.f1.name;
                Double score = value.f0.score;

                return sid + "," + sname + "," + subjectName + "," + score;
            }
        });

        result.print();
    }

    // POJO类: JavaBean
    public static class Subject{
        private int id;
        private String name;

        public Subject() {}

        public Subject(int id, String name) {
            this.id = id;
            this.name = name;
        }

        public int getId() {
            return id;
        }

        public void setId(int id) {
            this.id = id;
        }

        public String getName() {
            return name;
        }

        public void setName(String name) {
            this.name = name;
        }

        @Override
        public String toString() {
            return "Subject{" +
                    "id=" + id +
                    ", name='" + name + '\'' +
                    '}';
        }
    }

    public static class Score{
        private int id;
        private String name;
        private int subjectID;
        private Double score;

        public Score() {}

        public Score(int id, String name, int subjectID, Double score) {
            this.id = id;
            this.name = name;
            this.subjectID = subjectID;
            this.score = score;
        }

        @Override
        public String toString() {
            return "Score{" +
                    "id=" + id +
                    ", name='" + name + '\'' +
                    ", subjectID=" + subjectID +
                    ", score=" + score +
                    '}';
        }

        public int getId() {
            return id;
        }

        public void setId(int id) {
            this.id = id;
        }

        public String getName() {
            return name;
        }

        public void setName(String name) {
            this.name = name;
        }

        public int getSubjectID() {
            return subjectID;
        }

        public void setSubjectID(int subjectID) {
            this.subjectID = subjectID;
        }

        public Double getScore() {
            return score;
        }

        public void setScore(Double score) {
            this.score = score;
        }
    }
}

Union

将多个DataSet合并成一个DataSet

【注意】：union合并的DataSet的类型必须是一致的

示例
将以下数据进行取并集操作

数据集1

"hadoop", "hive", "flume"

数据集2

"hadoop", "hive", "spark"

步骤

• 构建批处理运行环境
• 使用 fromCollection 创建两个数据源
• 使用 union 将两个数据源关联在一起
• 打印测试

注意：union可以取并集，但是不会去重。

package batch.transformation;

import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;

public class UnionDemo {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        DataSource<String> source1 = env.fromElements("hadoop", "spark", "hive");
        DataSource<String> source2 = env.fromElements("yarn", "flink", "hive");

        source1.union(source2).print();
        /*
        Union算子 会进行合并, 不会进行重复判断
        Union算子 必须进行 同类型元素的合并, 哪怕是顶级类Object也不行, 必须是实体类(撇除继承关系)的类型一致才可以
         */
    }
}

Rebalance

Flink也有数据倾斜的时候，比如当前有数据量大概10亿条数据需要处理，在处理过程中可能会发生如图所示的状况：

这个时候本来总体数据量只需要10分钟解决的问题，出现了数据倾斜，机器1上的任务需要4个小时才能完成，那么其他3台机器执行完毕也要等待机器1执行完毕后才算整体将任务完成；
所以在实际的工作中，出现这种情况比较好的解决方案就是本节课要讲解的—rebalance

步骤

• 构建批处理运行环境
• 使用 env.generateSequence 创建0-100的并行数据
• 使用 fiter 过滤出来 大于8 的数字
• 使用map操作传入 RichMapFunction ，将当前子任务的ID和数字构建成一个元组
• 在RichMapFunction中可以使用getRuntimeContext.getIndexOfThisSubtask 获取子任务序号
• 打印测试

举例

在不使用rebalance的情况下，观察每一个线程执行的任务特点

package batch.transformation;

import org.apache.flink.api.common.functions.FilterFunction;
import org.apache.flink.api.common.functions.RichMapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.FilterOperator;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * @author lwh
 * @date 2023/4/12
 * @description 在不使用rebalance的情况下，观察每一个线程执行的任务特点
 **/
public class BatchDemoRebalance {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
        DataSource<Long> ds = env.generateSequence(0, 100);
        FilterOperator<Long> filter = ds.filter(new FilterFunction<Long>() {
            @Override
            public boolean filter(Long aLong) throws Exception {
                return aLong > 8;
            }
        });
        MapOperator<Long, Tuple2<Integer, Long>> countsInPartition = filter.map(new RichMapFunction<Long, Tuple2<Integer, Long>>() {
            @Override
            public Tuple2<Integer, Long> map(Long in) throws Exception {
                //获取并行时子任务的编号getRuntimeContext.getIndexOfThisSubtask
                return Tuple2.of(getRuntimeContext().getIndexOfThisSubtask(), in);
            }
        });
        countsInPartition.print();


    }
}

使用rebalance

package batch.transformation;

import org.apache.flink.api.common.functions.FilterFunction;
import org.apache.flink.api.common.functions.RichMapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.FilterOperator;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.operators.PartitionOperator;
import org.apache.flink.api.java.tuple.Tuple2;

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

        /**
         * 生成一个序列, 接受2个参数
         * 参数1: 开始
         * 参数2: 结束
         */
        DataSource<Long> source = env.generateSequence(1, 100);

        FilterOperator<Long> data = source.filter(new FilterFunction<Long>() {
            @Override
            public boolean filter(Long value) throws Exception {
                return value > 16;
            }
        });

        // 调用rebalance进行重新平衡
        PartitionOperator<Long> rebalancedData = data.rebalance();


        /**
         * RichMapFunction 是一个增强版的MapFunction
         * 增强了2个主要功能:
         * 1. 可以在里面获得运行时上下文环境"RuntimeContext", 通过getRuntimeContext获取
         * 2. 它带有open(构建类执行一次)和close(关闭类执行一次)的方法, 可以被复写
         */
        MapOperator<Long, Tuple2<Long, Long>> result = rebalancedData.map(new RichMapFunction<Long, Tuple2<Long, Long>>() {
            @Override
            public Tuple2<Long, Long> map(Long value) throws Exception {
                return Tuple2.of(value, getRuntimeContext().getIndexOfThisSubtask() + 0L);
            }
        });

        result.print();
    }
}

分区

partitionByHash

按照指定的key进行hash分区
分区数量和并行度有关，如果不设置并行度，会自动根据内容自动设置分区数量

还有一个同类函数：partitionByRange 按照key的范围进行排序

Hash和Range是Flink自行控制，我们无法控制
Hash规则是一样的key放入一个分区
Range是值范围在一个区域内（接近）的key，在一个分区

步骤

• 构建批处理运行环境
• 设置并行度为 2
• 使用 fromCollection 构建测试数据集
• 使用 partitionByHash 按照字符串的hash进行分区
• 调用 writeAsText 写入文件到 data/parition_output 目录中
• 打印测试

package batch.transformation;

import org.apache.flink.api.common.functions.RichMapFunction;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.operators.PartitionOperator;
import org.apache.flink.api.java.tuple.Tuple2;

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

        env.setParallelism(3);
        DataSource<Tuple2<Integer, Integer>> source = env.fromElements(
                Tuple2.of(1, 1),
                Tuple2.of(2, 1),
                Tuple2.of(3, 1),
                Tuple2.of(4, 1),
                Tuple2.of(5, 1),
                Tuple2.of(6, 1),
                Tuple2.of(7, 1),
                Tuple2.of(8, 1),
                Tuple2.of(9, 1),
                Tuple2.of(10, 1),
                Tuple2.of(11, 1),
                Tuple2.of(12, 1),
                Tuple2.of(13, 1),
                Tuple2.of(14, 1),
                Tuple2.of(15, 1)
        );

        /*
        partitionByHash 相同的key会在同一个分区内
        partitionByRange 按照分区字段的值 来进行均分范围, 相近值的数据 会在一个分区内
        range的计算是, 最小值和最大值之间, 按照并行度(分区数)最`范围`的均分
        range是字典值(ASCII)
         */
       PartitionOperator<Tuple2<Integer, Integer>> partitioned = source.partitionByHash(0);
//         PartitionOperator<Tuple2<Integer, Integer>> partitioned = source.partitionByRange(0);

        // 想自定义的话, 需要自定义分区逻辑
//        source.partitionCustom()

        partitioned.map(new RichMapFunction<Tuple2<Integer, Integer>, Tuple2<Integer, Long>>() {
            @Override
            public Tuple2<Integer, Long> map(Tuple2<Integer, Integer> value) throws Exception {
                return Tuple2.of(value.f0, getRuntimeContext().getIndexOfThisSubtask() + 0L);
            }
        }).print();
    }
}

sortPartition

根据指定的字段值进行分区的排序；

sortPartition(field, order)

package batch.transformation;

import org.apache.flink.api.common.functions.RichMapFunction;
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.operators.DataSource;
import org.apache.flink.api.java.operators.MapOperator;
import org.apache.flink.api.java.operators.SortPartitionOperator;
import org.apache.flink.api.java.tuple.Tuple2;

/**
 * @author lwh
 * @date 2023/4/12
 * @description
 **/
public class SortPartitionDemo {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        DataSource<Tuple2<String, Integer>> source = env.fromElements(
                Tuple2.of("hadoop", 11),
                Tuple2.of("hadoop", 21),
                Tuple2.of("hadoop", 3),
                Tuple2.of("hadoop", 16),
                Tuple2.of("hive", 13),
                Tuple2.of("hive", 31),
                Tuple2.of("hive", 21),
                Tuple2.of("hive", 11),
                Tuple2.of("hive", 15),
                Tuple2.of("hive", 19),
                Tuple2.of("spark", 51),
                Tuple2.of("spark", 61),
                Tuple2.of("spark", 19),
                Tuple2.of("spark", 35),
                Tuple2.of("spark", 66),
                Tuple2.of("spark", 76),
                Tuple2.of("flink", 11),
                Tuple2.of("flink", 51),
                Tuple2.of("flink", 31)
        );

        // 仅按照单词排序
        SortPartitionOperator<Tuple2<String, Integer>> sorted1 = source.sortPartition(0, Order.ASCENDING);

        MapOperator<Tuple2<String, Integer>, Tuple2<Integer, Tuple2<String, Integer>>> pt1 = pt(sorted1);
        pt1.print();

        // 按照单词以及数字排序
        System.out.println("-----------");
        SortPartitionOperator<Tuple2<String, Integer>> sorted2 = source.sortPartition(0, Order.ASCENDING).sortPartition(1, Order.ASCENDING);
        pt(sorted2).print();

        // 在分区内部按照单词排序
        System.out.println("-----------");
        pt(source.partitionByHash(0).sortPartition(0, Order.ASCENDING)).print();

        // 在分区内部按照单词和数字排序
        System.out.println("-----------");
        pt(source.partitionByRange(0).sortPartition(0, Order.ASCENDING).sortPartition(1, Order.ASCENDING)).print();

    }

    public static MapOperator<Tuple2<String, Integer>, Tuple2<Integer, Tuple2<String, Integer>>> pt(DataSet<Tuple2<String, Integer>> ds){
        return ds.map(new RichMapFunction<Tuple2<String, Integer>, Tuple2<Integer, Tuple2<String, Integer>>>() {
            @Override
            public Tuple2<Integer, Tuple2<String, Integer>> map(Tuple2<String, Integer> value) throws Exception {
                return Tuple2.of(getRuntimeContext().getIndexOfThisSubtask(), value);
            }
        });
    }

}