@Override
        public void forEachRemaining(Consumer<? super T> action) {
            if (action == null) throw new NullPointerException();
            Iterator<? extends T> i;
            if ((i = it) == null) {
                i = it = collection.iterator();
                est = (long)collection.size();
            }
            i.forEachRemaining(action);
        }

        @Override
        public boolean tryAdvance(Consumer<? super T> action) {
            if (action == null) throw new NullPointerException();
            if (it == null) {
                it = collection.iterator();
                est = (long) collection.size();
            }
            if (it.hasNext()) {
                action.accept(it.next());
                return true;
            }
            return false;
        }

整体流程就是这样。回顾一下：

• Collector 定义了必要的聚合操作函数；
• ReduceOps.makeRef 将 Collector 封装成一个计算对象 ReduceOps ，依赖的 ReducingSink 定义了具体的流数据消费过程；
• Spliterator 用于对流中的元素进行分区和遍历以及施加Sink指定的操作。

Pipeline

那么，Spliterator 又是从哪里来的呢？是通过类 java.util.stream.AbstractPipeline 的方法 sourceSpliterator 拿到的：

private Spliterator<?> sourceSpliterator(int terminalFlags) {
        // Get the source spliterator of the pipeline
        Spliterator<?> spliterator = null;
        if (sourceStage.sourceSpliterator != null) {
            spliterator = sourceStage.sourceSpliterator;
            sourceStage.sourceSpliterator = null;
        }
        else if (sourceStage.sourceSupplier != null) {
            spliterator = (Spliterator<?>) sourceStage.sourceSupplier.get();
            sourceStage.sourceSupplier = null;
        }
        else {
            throw new IllegalStateException(MSG_CONSUMED);
        }
        // code for isParallel
       return spliterator;
}

这里的 sourceStage 是一个 AbstractPipeline。 Pipeline 是实现流式计算的流水线抽象，也是Stream的实现类。可以看到，java.util.stream 定义了四种 pipeline: DoublePipeline, IntPipeline, LongPipeline, ReferencePipeline。可以重点看 ReferencePipeline 的实现。比如 filter, map

abstract class ReferencePipeline<P_IN, P_OUT>
        extends AbstractPipeline<P_IN, P_OUT, Stream<P_OUT>>
        implements Stream<P_OUT>

@Override
    public final Stream<P_OUT> filter(Predicate<? super P_OUT> predicate) {
        Objects.requireNonNull(predicate);
        return new StatelessOp<P_OUT, P_OUT>(this, StreamShape.REFERENCE,
                                     StreamOpFlag.NOT_SIZED) {
            @Override
            Sink<P_OUT> opWrapSink(int flags, Sink<P_OUT> sink) {
                return new Sink.ChainedReference<P_OUT, P_OUT>(sink) {
                    @Override
                    public void begin(long size) {
                        downstream.begin(-1);
                    }

                    @Override
                    public void accept(P_OUT u) {
                        if (predicate.test(u))
                            downstream.accept(u);
                    }
                };
            }
        };
    }

    @Override
    @SuppressWarnings("unchecked")
    public final <R> Stream<R> map(Function<? super P_OUT, ? extends R> mapper) {
        Objects.requireNonNull(mapper);
        return new StatelessOp<P_OUT, R>(this, StreamShape.REFERENCE,
                                     StreamOpFlag.NOT_SORTED | StreamOpFlag.NOT_DISTINCT) {
            @Override
            Sink<P_OUT> opWrapSink(int flags, Sink<R> sink) {
                return new Sink.ChainedReference<P_OUT, R>(sink) {
                    @Override
                    public void accept(P_OUT u) {
                        downstream.accept(mapper.apply(u));
                    }
                };
            }
        };
    }

套路基本一样，关键点在于 accept 方法。filter 只在满足条件时将值传给下一个 pipeline, 而 map 将计算的值传给下一个 pipeline. StatelessOp 没有什么逻辑，JDK文档解释是：Base class for a stateless intermediate stage of a Stream。相应还有一个 StatefulOp, Head。 这些都是 ReferencePipeline ，负责将值在 pipeline 之间传递，交给 Sink 去计算。