java stream distinct源码分析

首先看下两个简短的代码：

代码1：
@Test
public void stringTest(){
    List<String> list = new ArrayList<>();
    list.add(new String("hello"));
    list.add("world");
    list.add("hello");
    list.stream().distinct().forEach(System.out::println);
}

输出结果：

hello

world

代码2：

@Test
public void testClass(){
    List<A> list = new ArrayList<>();
    list.add(new A("hello"));
    list.add(new A("world"));
    list.add(new A("hello"));
    list.stream().distinct().forEach(t->System.out.println(t.getName()));
}

@Getter
public  class A{
    private String name;

    public A(String name){
        this.name = name;
    }
}

如果类A没有重写equals以及hashcode方法时，输出结果如下：

hello

world

hello

如果类A定义如下：

@Getter
public  class A{
    private String name;

    public A(String name){
        this.name = name;
    }

    @Override
    public int hashCode() {
        return name.hashCode();
    }

    @Override
    public boolean equals(Object obj) {
        if(obj = this){
            return true; 
        }
        if(obj == null || !(obj instanceof A)){
            return false;
        }
        return this.name.equals(((A) obj).getName());
    }
}

输出结果如下：

hello

world

具体为什么会有上述结果呢，让我们来看下DistinctOps源码：

static <T> ReferencePipeline<T, T> makeRef(AbstractPipeline<?, T, ?> upstream) {
    return new ReferencePipeline.StatefulOp<T, T>(upstream, StreamShape.REFERENCE,
                                                  StreamOpFlag.IS_DISTINCT | StreamOpFlag.NOT_SIZED) {

        <P_IN> Node<T> reduce(PipelineHelper<T> helper, Spliterator<P_IN> spliterator) {
            // If the stream is SORTED then it should also be ORDERED so the following will also
            // preserve the sort order
            TerminalOp<T, LinkedHashSet<T>> reduceOp
                    = ReduceOps.<T, LinkedHashSet<T>>makeRef(LinkedHashSet::new, LinkedHashSet::add,
                                                             LinkedHashSet::addAll);
            return Nodes.node(reduceOp.evaluateParallel(helper, spliterator));
        }

        @Override
        <P_IN> Node<T> opEvaluateParallel(PipelineHelper<T> helper,
                                          Spliterator<P_IN> spliterator,
                                          IntFunction<T[]> generator) {
            if (StreamOpFlag.DISTINCT.isKnown(helper.getStreamAndOpFlags())) {
                // No-op
                return helper.evaluate(spliterator, false, generator);
            }
            else if (StreamOpFlag.ORDERED.isKnown(helper.getStreamAndOpFlags())) {
                return reduce(helper, spliterator);
            }
            else {
                // Holder of null state since ConcurrentHashMap does not support null values
                AtomicBoolean seenNull = new AtomicBoolean(false);
                ConcurrentHashMap<T, Boolean> map = new ConcurrentHashMap<>();
                TerminalOp<T, Void> forEachOp = ForEachOps.makeRef(t -> {
                    if (t == null)
                        seenNull.set(true);
                    else
                        map.putIfAbsent(t, Boolean.TRUE);
                }, false);
                forEachOp.evaluateParallel(helper, spliterator);

                // If null has been seen then copy the key set into a HashSet that supports null values
                // and add null
                Set<T> keys = map.keySet();
                if (seenNull.get()) {
                    // TODO Implement a more efficient set-union view, rather than copying
                    keys = new HashSet<>(keys);
                    keys.add(null);
                }
                return Nodes.node(keys);
            }
        }

        @Override
        <P_IN> Spliterator<T> opEvaluateParallelLazy(PipelineHelper<T> helper, Spliterator<P_IN> spliterator) {
            if (StreamOpFlag.DISTINCT.isKnown(helper.getStreamAndOpFlags())) {
                // No-op
                return helper.wrapSpliterator(spliterator);
            }
            else if (StreamOpFlag.ORDERED.isKnown(helper.getStreamAndOpFlags())) {
                // Not lazy, barrier required to preserve order
                return reduce(helper, spliterator).spliterator();
            }
            else {
                // Lazy
                return new StreamSpliterators.DistinctSpliterator<>(helper.wrapSpliterator(spliterator));
            }
        }

        @Override
        Sink<T> opWrapSink(int flags, Sink<T> sink) {
            Objects.requireNonNull(sink);

            if (StreamOpFlag.DISTINCT.isKnown(flags)) {
                return sink;
            } else if (StreamOpFlag.SORTED.isKnown(flags)) {
                return new Sink.ChainedReference<T, T>(sink) {
                    boolean seenNull;
                    T lastSeen;

                    @Override
                    public void begin(long size) {
                        seenNull = false;
                        lastSeen = null;
                        downstream.begin(-1);
                    }

                    @Override
                    public void end() {
                        seenNull = false;
                        lastSeen = null;
                        downstream.end();
                    }

                    @Override
                    public void accept(T t) {
                        if (t == null) {
                            if (!seenNull) {
                                seenNull = true;
                                downstream.accept(lastSeen = null);
                            }
                        } else if (lastSeen == null || !t.equals(lastSeen)) {
                            downstream.accept(lastSeen = t);
                        }
                    }
                };
            } else {
                return new Sink.ChainedReference<T, T>(sink) {
                    Set<T> seen;

                    @Override
                    public void begin(long size) {
                        seen = new HashSet<>();
                        downstream.begin(-1);
                    }

                    @Override
                    public void end() {
                        seen = null;
                        downstream.end();
                    }

                    @Override
                    public void accept(T t) {
                        if (!seen.contains(t)) {
                            seen.add(t);
                            downstream.accept(t);
                        }
                    }
                };
            }
        }
    };
}

上面标浅蓝色部分就是原因，即java stream distinct底层是使用HashSet来实现去重处理的，HashSet本身又是基于HashMap来去重的，正如我们平时使用HashMap时需要保证HashMap的key必须重写equals以及hashcode方法，要想使用stream的distinct方法去重也必须保证涉及的类必须重写equals以及hashcode方法，否则就可能无法去重！！！