flink 计算留存 Java 版本_flink留存数据-优快云博客

本文链接：https://blog.youkuaiyun.com/cocoaxian/article/details/126756180
该代码示例展示了如何利用Apache Flink计算留存率。首先，它合并多个日文件数据，然后通过Flink API进行数据处理，包括去重、过滤、聚合等操作，以计算不同时间段的新增和留存用户。最后，输出留存率相关统计数据。
摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >
package demo.flink.retained;

import cn.hutool.core.date.DatePattern;
import cn.hutool.core.date.DateTime;
import cn.hutool.core.date.DateUtil;
import cn.hutool.core.io.FileUtil;
import cn.hutool.core.util.StrUtil;
import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.typeinfo.Types;
import org.apache.flink.api.java.DataSet;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.AggregateOperator;
import org.apache.flink.api.java.operators.FilterOperator;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.core.fs.FileSystem;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import demo.flink.wordcount.util.MappedBiggerFileReaderWrite;

import java.io.FileOutputStream;
import java.util.*;

/**
 * flink 计算留存
 * 数据源文件：每天一个文件
 *
 * @Author: lizixian
 * @date: 2022/9/7 23:43
 */
public class Retained4Demo {

    private static Logger logger = LoggerFactory.getLogger(Retained4Demo.class);

    private static String DATA_TEMP_PATH = "G:\\demo\\temp.txt";
    private static String DATA_BASE_PATH = "G:\\demo\\%s.txt";

    public static void main(String[] args) throws Exception {
        Map<String, List<String>> map = new TreeMap<>();
        Date date = DateUtil.parse("20220701");

        retained("20220701");

        System.out.println(map);
    }

    /**
     * 计算留存率
     *
     * @Author: lizixian
     * @date: 2022/7/8 22:30
     */
    public static List<String> retained(String day) throws Exception {

        logger.warn("计算 " + day + "这天的留存");

        if (!FileUtil.exist(String.format(DATA_BASE_PATH, day))) {
            logger.error(day + "这天 这天的文件不存在");
            return null;
        }

        // 合并临时文件
        FileUtil.del(DATA_TEMP_PATH);
        FileOutputStream fileOut = new FileOutputStream(DATA_TEMP_PATH, true);
        Date oldDay = DateUtil.parse("20150101");
        Date thatDay = DateUtil.parse(day);
        Date thatMonthFirst = DateUtil.offsetDay(thatDay, -29);
        while (oldDay.before(thatMonthFirst)) {
//            logger.warn("合并：" + oldDay.toLocaleString());
            String file = String.format(DATA_BASE_PATH, DateUtil.format(oldDay, DatePattern.PURE_DATE_FORMAT));
            oldDay = DateUtil.offsetDay(oldDay, 1);
            if (!FileUtil.exist(file)) {
                continue;
            }

            MappedBiggerFileReaderWrite reader = new MappedBiggerFileReaderWrite(file, 655360);
            while (reader.read() != -1) {
                fileOut.write(reader.getArray());
            }
            fileOut.write("\n".getBytes());
            reader.close();
        }
        fileOut.flush();
        fileOut.close();
        // 合并临时文件

        final ExecutionEnvironment env = ExecutionEnvironment.createLocalEnvironmentWithWebUI(new Configuration());
//        env.setParallelism(1);
        List<String> resultList = new ArrayList<>();

        // 从头读取全量设备，去重
        DataSet<Tuple3<String, String, Integer>> allData = env.readTextFile(DATA_TEMP_PATH).map(new Tokenizer(2)).distinct(0);
        logger.warn("allData Count= " + allData.count());

        DataSet<Tuple3<String, String, Integer>> union30Dya = null;
        for (int i = 29; i > 0; i--) {

            // 该日数据
            DateTime dateTime = DateUtil.offsetDay(thatDay, -i);
            String formatDay = DateUtil.format(dateTime, DatePattern.PURE_DATE_FORMAT);
            String itemDayFilePath = String.format(DATA_BASE_PATH, formatDay);

            if (!FileUtil.exist(itemDayFilePath)) {
                logger.error(formatDay + " 日的数据文件不存在");
                continue;
            }

            DataSet<Tuple3<String, String, Integer>> itemDayData = env.readTextFile(itemDayFilePath).map(new Tokenizer(1)).distinct(0);

            // 该日新增
            FilterOperator<Tuple3<String, String, Integer>> itemDayNewData = allData.union(itemDayData).groupBy(0).sum(2).filter(j -> j.f2 == 1);

            // 该日的留存数据
            if (union30Dya == null) {
                union30Dya = itemDayNewData;
            } else {
                union30Dya = union30Dya.union(itemDayNewData);
            }
        }

        // 聚合近30天的设备
        AggregateOperator<Tuple3<String, String, Integer>> sumMonth = union30Dya.groupBy(0).sum(2);
        logger.info("相较于历史，近30天内新增=" + sumMonth.count());

        // 这天之前的设备
        DataSet<Tuple3<String, String, Integer>> map = allData.union(sumMonth).map((MapFunction<Tuple3<String, String, Integer>, Tuple3<String, String, Integer>>) item -> {
            item.f2 = 2;
            return item;
        }).returns(Types.TUPLE(Types.STRING, Types.STRING, Types.INT)).distinct(0);
        logger.info(day + " 这天之前的设备数=" + map.count());

        // 该日数据
        String filePath = String.format(DATA_BASE_PATH, day);
        DataSet<Tuple3<String, String, Integer>> thatDayData = env.readTextFile(filePath).map(new Tokenizer(1)).distinct(0);
        logger.info("当天去重设备数=" + thatDayData.count());

        // 该日新增
        FilterOperator<Tuple3<String, String, Integer>> thatDayNewData = map.union(thatDayData).groupBy(0).sum(2).filter(i -> i.f2 == 1);
        resultList.add("1活=" + thatDayNewData.count());

        // 用30天内数据跟该日数据连接（）
        DataSet<Tuple3<String, String, Integer>> join = thatDayData.join(sumMonth).where(0).equalTo(0).map(i -> i.f1).returns(Types.TUPLE(Types.STRING, Types.STRING, Types.INT));

//        join.writeAsText( String.format(DATA_BASE_PATH, "sumMonth"), FileSystem.WriteMode.OVERWRITE).setParallelism(1);

        // 统计活跃数据
        long countTow = join.filter(i -> i.f2 == 1).count();
        long countThree = join.filter(i -> i.f2 == 2).count();
        long countFour = join.filter(i -> i.f2 == 3).count();
        long countFives = join.filter(i -> i.f2 == 4).count();
        long countSix = join.filter(i -> i.f2 == 5).count();

        resultList.add("2活= " + countTow);
        resultList.add("3活= " + countThree);
        resultList.add("4活= " + countFour);
        resultList.add("5活= " + countFives);
        resultList.add("6活= " + countSix);

        join.writeAsText(DATA_TEMP_PATH, FileSystem.WriteMode.OVERWRITE);
//        env.execute("留存统计" + System.currentTimeMillis());
        System.out.println(resultList);
        return resultList;

    }

    /**
     * 自定义分词器
     *
     * @Author: lizixian
     * @date: 2022/7/9 10:17
     */
    public static class Tokenizer implements MapFunction<String, Tuple3<String, String, Integer>> {

        private int initNum;

        public Tokenizer(int initNum) {
            this.initNum = initNum;
        }

        @Override
        public Tuple3<String, String, Integer> map(String s) throws Exception {
            if (StrUtil.isEmpty(s)) {
                return new Tuple3<>("", "", initNum);
            }
            List<String> split = StrUtil.split(s, "\t");
            if (split.size() == 1) {
                return new Tuple3<>(split.get(0), split.get(0), initNum);
            }
            return new Tuple3<>(split.get(0), split.get(1), initNum);
        }
    }

}