从ContinuousEventTimeTrigger/ContinuousProcessingTimeTrigger代码看如何实现一个自定义的触发器

背景

当我们想要实现提前触发计算的触发器时,我们可以使用ContinuousEventTimeTrigger/ContinuousProcessingTimeTrigger作为触发器达到比如几分钟触发一次计算并发送计算结果的类,我们本文就从代码角度解析下实现自定义触发器的一些注意事项

ContinuousEventTimeTrigger源码解析

@PublicEvolving
public class ContinuousEventTimeTrigger<W extends Window> extends Trigger<Object, W> {
    private static final long serialVersionUID = 1L;

    private final long interval;

    /** When merging we take the lowest of all fire timestamps as the new fire timestamp. */
    private final ReducingStateDescriptor<Long> stateDesc =
            new ReducingStateDescriptor<>("fire-time", new Min(), LongSerializer.INSTANCE);

    private ContinuousEventTimeTrigger(long interval) {
        this.interval = interval;
    }

    @Override
    public TriggerResult onElement(Object element, long timestamp, W window, TriggerContext ctx)
            throws Exception {

        if (window.maxTimestamp() <= ctx.getCurrentWatermark()) {
            // if the watermark is already past the window fire immediately
            // 这里只需要fire触发计算,为什么不是FIRE_AND_PURGE?
            return TriggerResult.FIRE;
        } else {
            // 这里注册一个结束窗口的计时器是否必要?
            ctx.registerEventTimeTimer(window.maxTimestamp());
        }

        ReducingState<Long> fireTimestamp = ctx.getPartitionedState(stateDesc);
        if (fireTimestamp.get() == null) {
            long start = timestamp - (timestamp % interval);
            long nextFireTimestamp = start + interval;
            ctx.registerEventTimeTimer(nextFireTimestamp);
            fireTimestamp.add(nextFireTimestamp);
        }

        return TriggerResult.CONTINUE;
    }

    @Override
    public TriggerResult onEventTime(long time, W window, TriggerContext ctx) throws Exception {

		//为什么是FIRE而不是FIRE_AND_PURGE
        if (time == window.maxTimestamp()) {
            return TriggerResult.FIRE;
        }

        ReducingState<Long> fireTimestampState = ctx.getPartitionedState(stateDesc);

        Long fireTimestamp = fireTimestampState.get();

        if (fireTimestamp != null && fireTimestamp == time) {
            fireTimestampState.clear();
            fireTimestampState.add(time + interval);
            ctx.registerEventTimeTimer(time + interval);
            return TriggerResult.FIRE;
        }

        return TriggerResult.CONTINUE;
    }

    @Override
    public TriggerResult onProcessingTime(long time, W window, TriggerContext ctx)
            throws Exception {
        return TriggerResult.CONTINUE;
    }

    @Override
    public void clear(W window, TriggerContext ctx) throws Exception {
        // 清除计时器
        ReducingState<Long> fireTimestamp = ctx.getPartitionedState(stateDesc);
        Long timestamp = fireTimestamp.get();
        if (timestamp != null) {
            ctx.deleteEventTimeTimer(timestamp);
            fireTimestamp.clear();
        }
    }

    @Override
    public boolean canMerge() {
        return true;
    }

    @Override
    public void onMerge(W window, OnMergeContext ctx) throws Exception {
        ctx.mergePartitionedState(stateDesc);
        Long nextFireTimestamp = ctx.getPartitionedState(stateDesc).get();
        if (nextFireTimestamp != null) {
            ctx.registerEventTimeTimer(nextFireTimestamp);
        }
    }

    @Override
    public String toString() {
        return "ContinuousEventTimeTrigger(" + interval + ")";
    }

    @VisibleForTesting
    public long getInterval() {
        return interval;
    }

    /**
     * Creates a trigger that continuously fires based on the given interval.
     *
     * @param interval The time interval at which to fire.
     * @param  The type of {@link Window Windows} on which this trigger can operate.
     */
    public static <W extends Window> ContinuousEventTimeTrigger<W> of(Time interval) {
        return new ContinuousEventTimeTrigger<>(interval.toMilliseconds());
    }

    private static class Min implements ReduceFunction<Long> {
        private static final long serialVersionUID = 1L;

        @Override
        public Long reduce(Long value1, Long value2) throws Exception {
            return Math.min(value1, value2);
        }
    }
}

ContinuousProcessingTimeTrigger源码

@PublicEvolving
public class ContinuousProcessingTimeTrigger<W extends Window> extends Trigger<Object, W> {
    private static final long serialVersionUID = 1L;

    private final long interval;

    /** When merging we take the lowest of all fire timestamps as the new fire timestamp. */
    private final ReducingStateDescriptor<Long> stateDesc =
            new ReducingStateDescriptor<>("fire-time", new Min(), LongSerializer.INSTANCE);

    private ContinuousProcessingTimeTrigger(long interval) {
        this.interval = interval;
    }

    @Override
    public TriggerResult onElement(Object element, long timestamp, W window, TriggerContext ctx)
            throws Exception {
        ReducingState<Long> fireTimestamp = ctx.getPartitionedState(stateDesc);

        timestamp = ctx.getCurrentProcessingTime();
		// 注册计时器,为什么这里不需要类似ContinuousEventTimeTrigger一样注册一个窗口结束时间的计时器?
        if (fireTimestamp.get() == null) {
            long start = timestamp - (timestamp % interval);
            long nextFireTimestamp = start + interval;

            ctx.registerProcessingTimeTimer(nextFireTimestamp);

            fireTimestamp.add(nextFireTimestamp);
            return TriggerResult.CONTINUE;
        }
        return TriggerResult.CONTINUE;
    }

    @Override
    public TriggerResult onEventTime(long time, W window, TriggerContext ctx) throws Exception {
        return TriggerResult.CONTINUE;
    }

    @Override
    public TriggerResult onProcessingTime(long time, W window, TriggerContext ctx)
            throws Exception {
        ReducingState<Long> fireTimestamp = ctx.getPartitionedState(stateDesc);

        if (fireTimestamp.get().equals(time)) {
            fireTimestamp.clear();
            fireTimestamp.add(time + interval);
            ctx.registerProcessingTimeTimer(time + interval);
            return TriggerResult.FIRE;
        }
        //为什么这里没有FIRE_AND_PURGE?状态是何时清理的?
        return TriggerResult.CONTINUE;
    }

    @Override
    public void clear(W window, TriggerContext ctx) throws Exception {
        //清除计时器
        // State could be merged into new window.
        ReducingState<Long> fireTimestamp = ctx.getPartitionedState(stateDesc);
        Long timestamp = fireTimestamp.get();
        if (timestamp != null) {
            ctx.deleteProcessingTimeTimer(timestamp);
            fireTimestamp.clear();
        }
    }

    @Override
    public boolean canMerge() {
        return true;
    }

    @Override
    public void onMerge(W window, OnMergeContext ctx) throws Exception {
        // States for old windows will lose after the call.
        ctx.mergePartitionedState(stateDesc);

        // Register timer for this new window.
        Long nextFireTimestamp = ctx.getPartitionedState(stateDesc).get();
        if (nextFireTimestamp != null) {
            ctx.registerProcessingTimeTimer(nextFireTimestamp);
        }
    }

    @VisibleForTesting
    public long getInterval() {
        return interval;
    }

    @Override
    public String toString() {
        return "ContinuousProcessingTimeTrigger(" + interval + ")";
    }

    /**
     * Creates a trigger that continuously fires based on the given interval.
     *
     * @param interval The time interval at which to fire.
     * @param  The type of {@link Window Windows} on which this trigger can operate.
     */
    public static <W extends Window> ContinuousProcessingTimeTrigger<W> of(Time interval) {
        return new ContinuousProcessingTimeTrigger<>(interval.toMilliseconds());
    }

    private static class Min implements ReduceFunction<Long> {
        private static final long serialVersionUID = 1L;

        @Override
        public Long reduce(Long value1, Long value2) throws Exception {
            return Math.min(value1, value2);
        }
    }
}

疑问:

1.为什么ContinuousEventTimeTrigger需要注册一个窗口结束时间的计时器而ContinuousProcessingTimeTrigger不注册?
答案: 其实我们需要看下它注册后的目的作用是什么,ContinuousEventTimeTrigger的ontimer在处理窗口结束的触发器时会返回FIRE触发计算,那问题就来了,如果只是触发计算,那么如果没有,那么仅仅只是窗口结束的时候没有触发一次计算而已。所以这里不是必须的
2.为什么ContinuousEventTimeTrigger/ContinuousProcessingTimeTrigger返回的结果是FIRE而不是FIRE_AND_PURGE?状态是什么时候清理的?
答案: 首先要明确状态的清理这个逻辑,状态其实包括窗口的状态,触发器的状态等,返回FIRE仅仅是触发计算,而不会清理任何状态,而假设返回FIRE_AND_PURGE的作用是触发计算,并进行窗口状态的清理(注意这里是不包括触发器的状态的清理的),其实状态的清理是由WindowOperator在清理时间到时进行的(对于事件时间是窗口结束时间+迟到容忍间隔时间,对于处理时间是窗口结束时间),所以不必要在窗口结束时间到的时候返回FIRE_AND_PURGE,可以统一由WindowOperator在清理时间到之后统一清理状态

你可能感兴趣的:(flink,大数据,flink)