Module java.base
Package java.util

Interface Spliterator<T>

Type Parameters:
T - the type of elements returned by this Spliterator
All Known Subinterfaces:
Spliterator.OfDouble, Spliterator.OfInt, Spliterator.OfLong, Spliterator.OfPrimitive<T,T_CONS,T_SPLITR>
All Known Implementing Classes:
Spliterators.AbstractDoubleSpliterator, Spliterators.AbstractIntSpliterator, Spliterators.AbstractLongSpliterator, Spliterators.AbstractSpliterator
public interface Spliterator<T>
An object for traversing and partitioning elements of a source. The source of elements covered by a Spliterator could be, for example, an array, a Collection, an IO channel, or a generator function.

A Spliterator may traverse elements individually (tryAdvance()) or sequentially in bulk (forEachRemaining()).

A Spliterator may also partition off some of its elements (using trySplit()) as another Spliterator, to be used in possibly-parallel operations. Operations using a Spliterator that cannot split, or does so in a highly imbalanced or inefficient manner, are unlikely to benefit from parallelism. Traversal and splitting exhaust elements; each Spliterator is useful for only a single bulk computation.

A Spliterator also reports a set of characteristics() of its structure, source, and elements from among ORDERED, DISTINCT, SORTED, SIZED, NONNULL, IMMUTABLE, CONCURRENT, and SUBSIZED. These may be employed by Spliterator clients to control, specialize or simplify computation. For example, a Spliterator for a Collection would report SIZED, a Spliterator for a Set would report DISTINCT, and a Spliterator for a SortedSet would also report SORTED. Characteristics are reported as a simple unioned bit set. Some characteristics additionally constrain method behavior; for example if ORDERED, traversal methods must conform to their documented ordering. New characteristics may be defined in the future, so implementors should not assign meanings to unlisted values.

A Spliterator that does not report IMMUTABLE or CONCURRENT is expected to have a documented policy concerning: when the spliterator binds to the element source; and detection of structural interference of the element source detected after binding. A late-binding Spliterator binds to the source of elements at the point of first traversal, first split, or first query for estimated size, rather than at the time the Spliterator is created. A Spliterator that is not late-binding binds to the source of elements at the point of construction or first invocation of any method. Modifications made to the source prior to binding are reflected when the Spliterator is traversed. After binding a Spliterator should, on a best-effort basis, throw ConcurrentModificationException if structural interference is detected. Spliterators that do this are called fail-fast. The bulk traversal method (forEachRemaining()) of a Spliterator may optimize traversal and check for structural interference after all elements have been traversed, rather than checking per-element and failing immediately.

Spliterators can provide an estimate of the number of remaining elements via the estimateSize() method. Ideally, as reflected in characteristic SIZED, this value corresponds exactly to the number of elements that would be encountered in a successful traversal. However, even when not exactly known, an estimated value may still be useful to operations being performed on the source, such as helping to determine whether it is preferable to split further or traverse the remaining elements sequentially.

Despite their obvious utility in parallel algorithms, spliterators are not expected to be thread-safe; instead, implementations of parallel algorithms using spliterators should ensure that the spliterator is only used by one thread at a time. This is generally easy to attain via serial thread-confinement, which often is a natural consequence of typical parallel algorithms that work by recursive decomposition. A thread calling trySplit() may hand over the returned Spliterator to another thread, which in turn may traverse or further split that Spliterator. The behaviour of splitting and traversal is undefined if two or more threads operate concurrently on the same spliterator. If the original thread hands a spliterator off to another thread for processing, it is best if that handoff occurs before any elements are consumed with tryAdvance(), as certain guarantees (such as the accuracy of estimateSize() for SIZED spliterators) are only valid before traversal has begun.

Primitive subtype specializations of Spliterator are provided for int, long, and double values. The subtype default implementations of tryAdvance(java.util.function.Consumer) and forEachRemaining(java.util.function.Consumer) box primitive values to instances of their corresponding wrapper class. Such boxing may undermine any performance advantages gained by using the primitive specializations. To avoid boxing, the corresponding primitive-based methods should be used. For example, Spliterator.OfPrimitive.tryAdvance(java.util.function.IntConsumer) and Spliterator.OfPrimitive.forEachRemaining(java.util.function.IntConsumer) should be used in preference to Spliterator.OfInt.tryAdvance(java.util.function.Consumer) and Spliterator.OfInt.forEachRemaining(java.util.function.Consumer). Traversal of primitive values using boxing-based methods tryAdvance() and forEachRemaining() does not affect the order in which the values, transformed to boxed values, are encountered.

API Note:

Spliterators, like Iterators, are for traversing the elements of a source. The Spliterator API was designed to support efficient parallel traversal in addition to sequential traversal, by supporting decomposition as well as single-element iteration. In addition, the protocol for accessing elements via a Spliterator is designed to impose smaller per-element overhead than Iterator, and to avoid the inherent race involved in having separate methods for hasNext() and next().

For mutable sources, arbitrary and non-deterministic behavior may occur if the source is structurally interfered with (elements added, replaced, or removed) between the time that the Spliterator binds to its data source and the end of traversal. For example, such interference will produce arbitrary, non-deterministic results when using the java.util.stream framework.

Structural interference of a source can be managed in the following ways (in approximate order of decreasing desirability):

  • The source cannot be structurally interfered with.
    For example, an instance of CopyOnWriteArrayList is an immutable source. A Spliterator created from the source reports a characteristic of IMMUTABLE.
  • The source manages concurrent modifications.
    For example, a key set of a ConcurrentHashMap is a concurrent source. A Spliterator created from the source reports a characteristic of CONCURRENT.
  • The mutable source provides a late-binding and fail-fast Spliterator.
    Late binding narrows the window during which interference can affect the calculation; fail-fast detects, on a best-effort basis, that structural interference has occurred after traversal has commenced and throws ConcurrentModificationException. For example, ArrayList, and many other non-concurrent Collection classes in the JDK, provide a late-binding, fail-fast spliterator.
  • The mutable source provides a non-late-binding but fail-fast Spliterator.
    The source increases the likelihood of throwing ConcurrentModificationException since the window of potential interference is larger.
  • The mutable source provides a late-binding and non-fail-fast Spliterator.
    The source risks arbitrary, non-deterministic behavior after traversal has commenced since interference is not detected.
  • The mutable source provides a non-late-binding and non-fail-fast Spliterator.
    The source increases the risk of arbitrary, non-deterministic behavior since non-detected interference may occur after construction.

Example. Here is a class (not a very useful one, except for illustration) that maintains an array in which the actual data are held in even locations, and unrelated tag data are held in odd locations. Its Spliterator ignores the tags. 

 
 class TaggedArray<T> {
   private final Object[] elements; // immutable after construction
   TaggedArray(T[] data, Object[] tags) {
     int size = data.length;
     if (tags.length != size) throw new IllegalArgumentException();
     this.elements = new Object[2 * size];
     for (int i = 0, j = 0; i < size; ++i) {
       elements[j++] = data[i];
       elements[j++] = tags[i];
     }
   }

   public Spliterator<T> spliterator() {
     return new TaggedArraySpliterator<>(elements, 0, elements.length);
   }

   static class TaggedArraySpliterator<T> implements Spliterator<T> {
     private final Object[] array;
     private int origin; // current index, advanced on split or traversal
     private final int fence; // one past the greatest index

     TaggedArraySpliterator(Object[] array, int origin, int fence) {
       this.array = array; this.origin = origin; this.fence = fence;
     }

     public void forEachRemaining(Consumer<? super T> action) {
       for (; origin < fence; origin += 2)
         action.accept((T) array[origin]);
     }

     public boolean tryAdvance(Consumer<? super T> action) {
       if (origin < fence) {
         action.accept((T) array[origin]);
         origin += 2;
         return true;
       }
       else // cannot advance
         return false;
     }

     public Spliterator<T> trySplit() {
       int lo = origin; // divide range in half
       int mid = ((lo + fence) >>> 1) & ~1; // force midpoint to be even
       if (lo < mid) { // split out left half
         origin = mid; // reset this Spliterator's origin
         return new TaggedArraySpliterator<>(array, lo, mid);
       }
       else       // too small to split
         return null;
     }

     public long estimateSize() {
       return (long)((fence - origin) / 2);
     }

     public int characteristics() {
       return ORDERED | SIZED | IMMUTABLE | SUBSIZED;
     }
   }
 }

As an example how a parallel computation framework, such as the java.util.stream package, would use Spliterator in a parallel computation, here is one way to implement an associated parallel forEach, that illustrates the primary usage idiom of splitting off subtasks until the estimated amount of work is small enough to perform sequentially. Here we assume that the order of processing across subtasks doesn't matter; different (forked) tasks may further split and process elements concurrently in undetermined order. This example uses a CountedCompleter; similar usages apply to other parallel task constructions. 


 static <T> void parEach(TaggedArray<T> a, Consumer<T> action) {
   Spliterator<T> s = a.spliterator();
   long targetBatchSize = s.estimateSize() / (ForkJoinPool.getCommonPoolParallelism() * 8);
   new ParEach(null, s, action, targetBatchSize).invoke();
 }

 static class ParEach<T> extends CountedCompleter<Void> {
   final Spliterator<T> spliterator;
   final Consumer<T> action;
   final long targetBatchSize;

   ParEach(ParEach<T> parent, Spliterator<T> spliterator,
           Consumer<T> action, long targetBatchSize) {
     super(parent);
     this.spliterator = spliterator; this.action = action;
     this.targetBatchSize = targetBatchSize;
   }

   public void compute() {
     Spliterator<T> sub;
     while (spliterator.estimateSize() > targetBatchSize &&
            (sub = spliterator.trySplit()) != null) {
       addToPendingCount(1);
       new ParEach<>(this, sub, action, targetBatchSize).fork();
     }
     spliterator.forEachRemaining(action);
     propagateCompletion();
   }
 }

Implementation Note:
If the boolean system property org.openjdk.java.util.stream.tripwire is set to true then diagnostic warnings are reported if boxing of primitive values occur when operating on primitive subtype specializations.
Since:
1.8
See Also: