(2)Flink CEP SQL严格近邻代码演示-风控系统构建利器

上一篇我们对Flink CEP做了简单介绍,这一篇我们通过代码来演示一下Flink CEP SQL中的严格近邻效果:
(1)pom依赖:
<dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-cep_${scala.binary.version}</artifactId>
    <version>${flink.version}</version>
</dependency>
(2)定义一个消息对象
public static class Ticker {
    public long id;
    public String symbol;
    public long price;
    public long tax;
    public LocalDateTime rowtime;

    public Ticker() {
    }

    public Ticker(long id, String symbol, long price, long item, LocalDateTime rowtime) {
        this.id = id;
        this.symbol = symbol;
        this.price = price;
        this.tax = tax;
        this.rowtime = rowtime;
    }
}
(3)构造数据,定义事件组合
public static void main(String[] args) {
    EnvironmentSettings settings = null;
    StreamTableEnvironment tEnv = null;
    try {

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        settings = EnvironmentSettings.newInstance()
                .useBlinkPlanner()
                .inStreamingMode()
                .build();
        tEnv = StreamTableEnvironment.create(env, settings);
        System.out.println("===============CEP_SQL_9=================");
        final DateTimeFormatter dateTimeFormatter = DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss");
        DataStream<Ticker> dataStream =
            env.fromElements(
                new Ticker(1, "ACME", 22, 1, LocalDateTime.parse("2021-12-10 10:00:00", dateTimeFormatter)),
                new Ticker(3, "ACME", 19, 1, LocalDateTime.parse("2021-12-10 10:00:02", dateTimeFormatter)),
                new Ticker(4, "ACME", 23, 3, LocalDateTime.parse("2021-12-10 10:00:03", dateTimeFormatter)),
                new Ticker(5, "Apple", 25, 2, LocalDateTime.parse("2021-12-10 10:00:04", dateTimeFormatter)),
                new Ticker(6, "Apple", 18, 1, LocalDateTime.parse("2021-12-10 10:00:05", dateTimeFormatter)),
                new Ticker(7, "Apple", 16, 1, LocalDateTime.parse("2021-12-10 10:00:06", dateTimeFormatter)),
                new Ticker(8, "Apple", 14, 2, LocalDateTime.parse("2021-12-10 10:00:07", dateTimeFormatter)),
                new Ticker(9, "Apple", 15, 2, LocalDateTime.parse("2021-12-10 10:00:08", dateTimeFormatter)),
                new Ticker(10, "Apple", 25, 2, LocalDateTime.parse("2021-12-10 10:00:09", dateTimeFormatter)),
                new Ticker(11, "Apple", 22, 1, LocalDateTime.parse("2021-12-10 10:00:11", dateTimeFormatter)),
                new Ticker(12, "Apple", 15, 1, LocalDateTime.parse("2021-12-10 10:00:12", dateTimeFormatter)),
                new Ticker(13, "Apple", 19, 1, LocalDateTime.parse("2021-12-10 10:00:13", dateTimeFormatter)),
                new Ticker(14, "Apple", 25, 1, LocalDateTime.parse("2021-12-10 10:00:14", dateTimeFormatter)),
                new Ticker(15, "Apple", 19, 1, LocalDateTime.parse("2021-12-10 10:00:15", dateTimeFormatter)),
                new Ticker(16, "Apple", 15, 1, LocalDateTime.parse("2021-12-10 10:00:16", dateTimeFormatter)),
                new Ticker(17, "Apple", 19, 1, LocalDateTime.parse("2021-12-10 10:00:17", dateTimeFormatter)),
                new Ticker(18, "Apple", 15, 1, LocalDateTime.parse("2021-12-10 10:00:18", dateTimeFormatter)));

        Table table = tEnv.fromDataStream(dataStream, Schema.newBuilder()
                .column("id", DataTypes.BIGINT())
                .column("symbol", DataTypes.STRING())
                .column("price", DataTypes.BIGINT())
                .column("tax", DataTypes.BIGINT())
                .column("rowtime", DataTypes.TIMESTAMP(3))
                .watermark("rowtime", "rowtime - INTERVAL '1' SECOND")
                .build());
        tEnv.createTemporaryView("CEP_SQL_9", table);

        String sql = "SELECT * " +
                "FROM CEP_SQL_9 " +
                "    MATCH_RECOGNIZE ( " +
                "        PARTITION BY symbol " +       //按symbol分区,将相同卡号的数据分到同一个计算节点上。
                "        ORDER BY rowtime " +          //在窗口内,对事件时间进行排序。
                "        MEASURES " +                   //定义如何根据匹配成功的输入事件构造输出事件
                "            e1.id as id,"+
                "            AVG(e1.price) as avgPrice,"+
                "            e1.rowtime AS start_tstamp, " +
                "            e3.rowtime AS end_tstamp " +
                "        ONE ROW PER MATCH " +                                      //匹配成功输出一条
                "        AFTER MATCH  skip to next row " +                   //匹配后跳转到下一行
                "        PATTERN ( e1 e2 e3) WITHIN INTERVAL '2' MINUTE" +
                "        DEFINE " +                                                 //定义各事件的匹配条件
                "            e1 AS " +
                "                e1.price = 25 , " +
                "            e2 AS " +
                "                e2.price > 10 ," +
                "            e3 AS " +
                "                e3.price = 15 " +
                "    ) MR";
        
        
        TableResult res = tEnv.executeSql(sql);
        res.print();
        tEnv.dropTemporaryView("CEP_SQL_9");
            } catch (Exception e) {
                LOG.error(e.getMessage(), e);
            }
}
(4)关键代码解释:
输出两分钟内匹配到的数据,输出信息:
(5)执行效果:

从数据集中匹配到了两组符合要求的数据。

 

posted @ 2022-08-13 11:10  NBI大数据可视化分析  阅读(56)  评论(0编辑  收藏  举报