14 ​​buffer​​ window​​ groupBy

本章系统讲解Reactor中三大批处理操作符：buffer、window**和groupBy**，通过示例对比（如buffer(3) vs window(3)）揭示其核心差异。核心要点如下：

二、批处理操作原理深度解析¶

1. 响应式流规范中的角色¶

• 动态资源管理：需实现Subscription动态调整数据流速率
• 边界触发机制：通过request(n)信号控制窗口/分组的生成节奏
• 错误传播规则：任一分组内发生错误会终止整个数据流

2. 数据流处理流程对比¶

Buffer操作：¶

sequenceDiagram
    participant Source as Flux
    participant Buffer as BufferSubscriber
    participant Subscriber as 实际消费者

    Source->>Buffer: 请求数据
    Buffer->>Source: 请求3个元素
    Source->>Buffer: 发射元素1,2,3
    Buffer->>Subscriber: 发射[1,2,3]
    Source->>Buffer: 发射元素4,5
    Buffer->>Subscriber: 发射[4,5]

sequenceDiagram
    participant Source as Flux
    participant Buffer as BufferSubscriber
    participant Subscriber as 实际消费者

    Source->>Buffer: 请求数据
    Buffer->>Source: 请求3个元素
    Source->>Buffer: 发射元素1,2,3
    Buffer->>Subscriber: 发射[1,2,3]
    Source->>Buffer: 发射元素4,5
    Buffer->>Subscriber: 发射[4,5]

Window操作：¶

sequenceDiagram
    participant Source as Flux
    participant Window as WindowSubscriber
    participant Subscriber as 实际消费者

    Source->>Window: 请求数据
    Window->>Source: 请求3个元素
    Source->>Window: 发射元素1,2,3
    Window->>Subscriber: 发射Flux[1,2,3]
    Source->>Window: 发射元素4,5
    Window->>Subscriber: 发射Flux[4,5]

sequenceDiagram
    participant Source as Flux
    participant Window as WindowSubscriber
    participant Subscriber as 实际消费者

    Source->>Window: 请求数据
    Window->>Source: 请求3个元素
    Source->>Window: 发射元素1,2,3
    Window->>Subscriber: 发射Flux[1,2,3]
    Source->>Window: 发射元素4,5
    Window->>Subscriber: 发射Flux[4,5]

GroupBy操作：¶

sequenceDiagram
    participant Source as Flux
    participant GroupBy as GroupBySubscriber
    participant Subscriber as 实际消费者

    Source->>GroupBy: 请求数据
    GroupBy->>Source: 请求元素
    Source->>GroupBy: 发射元素A(key1)
    GroupBy->>Subscriber: 发射GroupedFlux[key1]
    GroupBy->>GroupedFlux[key1]: 发射元素A

sequenceDiagram
    participant Source as Flux
    participant GroupBy as GroupBySubscriber
    participant Subscriber as 实际消费者

    Source->>GroupBy: 请求数据
    GroupBy->>Source: 请求元素
    Source->>GroupBy: 发射元素A(key1)
    GroupBy->>Subscriber: 发射GroupedFlux[key1]
    GroupBy->>GroupedFlux[key1]: 发射元素A

3. 关键机制¶

• 窗口触发策略：
• buffer：基于数量/时间触发
• window：基于数量/时间或子Flux完成事件
• 分组策略：
• groupBy：通过Function<? super T, ? extends K>生成分组键
• 支持动态分组（键值可随元素变化）

三、源码解读（Reactor 3.x）¶

1. Buffer操作实现¶

// FluxBuffer.java
public final class FluxBuffer<T> extends FluxOperator<T, List<T>> {
    public FluxBuffer(Flux<? extends T> source, int bufferSize) {
        super(source);
        this.bufferSize = bufferSize;
    }

    @Override
    public void subscribe(Subscriber<? super List<T>> subscriber) {
        source.subscribe(new BufferSubscriber<>(subscriber, bufferSize));
    }
}

// BufferSubscriber关键逻辑
void drain() {
    if (buffer.size() >= bufferSize) {
        List<T> emitted = buffer;
        buffer = new ArrayList<>();
        actual.onNext(emitted);
    }
}

2. Window操作实现¶

// FluxWindow.java
public final class FluxWindow<T> extends FluxOperator<T, Flux<T>> {
    public FluxWindow(Flux<? extends T> source, int windowSize) {
        super(source);
        this.windowSize = windowSize;
    }

    @Override
    public void subscribe(Subscriber<? super Flux<T>> subscriber) {
        source.subscribe(new WindowSubscriber<>(subscriber, windowSize));
    }
}

// WindowSubscriber关键逻辑
void windowComplete() {
    Flux<T> window = Flux.from(buffer);
    actual.onNext(window);
    buffer = new ArrayList<>();
}

3. GroupBy操作实现¶

// FluxGroupBy.java
public final class FluxGroupBy<T, K> extends Flux<GroupedFlux<K, T>> {
    public FluxGroupBy(Flux<? extends T> source, Function<? super T, ? extends K> keySelector) {
        super(source);
        this.keySelector = keySelector;
    }

    @Override
    public void subscribe(Subscriber<? super GroupedFlux<K, T>> subscriber) {
        source.subscribe(new GroupBySubscriber<>(subscriber, keySelector));
    }
}

// GroupBySubscriber关键逻辑
void onNext(T t) {
    K key = keySelector.apply(t);
    GroupedFlux<K, T> group = groups.get(key);
    if (group == null) {
        group = new GroupedFlux<>(key);
        groups.put(key, group);
        subscriber.onNext(group);
    }
    group.onNext(t);
}

四、与Java 8 Stream的对比分析¶

五、最佳实践与性能优化¶

1. 典型应用场景¶

// 实时日志分组统计
Flux<LogEvent> logs = ...
logs.groupBy(LogEvent::getUserId)
    .flatMap(group -> group.count())
    .subscribe(count -> System.out.println("User activity: " + count));

//