Question

在FLink CEP 1.7.1上进行开发时，我在使用“跳过策略”时遇到问题。这是我的代码，

public class CepSurf {
private static final Logger LOG = LoggerFactory.getLogger(CepSurf.class);

public static void main (String[] args) {
    StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime);
    env.setParallelism(1);
    DataStream<Tuple3<String, String, Long>> source = env.fromElements(Tuple3.of("b", "b1", 1L),
            Tuple3.of("b", "b2", 2L),
            Tuple3.of("b", "b3", 3L),
            Tuple3.of("c", "c", 4L));

    // pattern "b+ c"
    Pattern<Tuple3<String, String, Long>, ?> pattern = Pattern.<Tuple3<String, String, Long>>begin("pattern-b", AfterMatchSkipStrategy.skipToFirst("pattern-b")).where(new SimpleCondition<Tuple3<String, String, Long>>() {
        @Override
        public boolean filter(Tuple3<String, String, Long> value) throws Exception {
            if ("b".equals(value.f0)) {
                LOG.debug("Found b! {}", value);
                return true;
            }
            else
                return false;
        }
    }).oneOrMore().next("pattern-c").where(new SimpleCondition<Tuple3<String, String, Long>>() {
        @Override
        public boolean filter(Tuple3<String, String, Long> value) throws Exception {
            if ("c".equals(value.f0)) {
                LOG.debug("Found c! {}", value);
                return true;
            }
            else
                return false;
        }
    }).within(Time.seconds(1));

    PatternStream<Tuple3<String, String, Long>> patternStream = CEP.pattern(source.assignTimestampsAndWatermarks(new AssignerWithPunctuatedWatermarks<Tuple3<String, String, Long>>() {
        @Nullable
        @Override
        public Watermark checkAndGetNextWatermark(Tuple3<String, String, Long> lastElement, long extractedTimestamp) {
            return new Watermark(extractedTimestamp - 1000);
        }

        @Override
        public long extractTimestamp(Tuple3<String, String, Long> element, long previousElementTimestamp) {
            return element.f2;
        }
    }), pattern);

    patternStream.select(new PatternSelectFunction<Tuple3<String,String,Long>, String>() {
        @Override
        public String select(Map<String, List<Tuple3<String, String, Long>>> pattern) throws Exception {
            String phenomenon = pattern.get("pattern-b").stream().map(tuple -> tuple.f1).collect(Collectors.joining("-"));
            phenomenon = String.join("-", phenomenon, pattern.get("pattern-c").get(0).f1);
            return "MAYDAY" + phenomenon + "MAYDAY";
        }
    }).print();

    try {
        env.execute("orderness");
    } catch (Exception e) {
        e.printStackTrace();
    }
}

}

控制台输出如下所示，

2019-03-17 18:02:16 [nt to Std. Out (1/1):      8748] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b1,1)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8766] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b2,2)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8766] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b2,2)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8768] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b2,2)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8768] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b3,3)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8769] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b3,3)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8769] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b3,3)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8769] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b3,3)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8769] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:45] : Found b! (b,b3,3)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8771] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:55] : Found c! (c,c,4)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8774] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:55] : Found c! (c,c,4)
2019-03-17 18:02:16 [nt to Std. Out (1/1):      8776] DEBUG [com.jd.alchemist.demon.flink.examples.lab.CepSurf:55] : Found c! (c,c,4)
MAYDAYb1-b2-b3-cMAYDAY

有两个问题，

当我使用skipToFirst策略时，我期望结果为

MAYDAYb1-b2-b3-cMAYDAY
MAYDAYb2-b3-cMAYDAY
MAYDAYb3-cMAYDAY

但是我得到的MAYDAYb1-b2-b3-cMAYDAY与Flink official doc.不一致。我的代码有什么问题？

为什么filter对元组b1，b2，b3和c的调用时间分别是1、3、5和3？

非常感谢！

Flink CEP中的跳过策略

0 个答案: