stanfordnlp/CoreNLP · error · ArrayIndexOutOfBoundsException

Out of bounds: >=

Error message

Out of bounds: %d >= %d

What it means

ConcurrentHashIndex.get(int i) throws ArrayIndexOutOfBoundsException when the requested index i is not smaller than the index's current size (indexSize). The index maps items to dense integer IDs; asking for an ID that was never assigned (or belongs to a stale snapshot) is out of bounds. The message reports the requested index and the current bound.

Solutions

1. Check 0 <= i && i < index.size() before calling get(i)
2. Treat indexOf() == Integer-UNKNOWN (-1) as 'item absent' and skip the get()
3. Load the index from the same saved snapshot that produced the ids (saveToWriter/loadFromReader pair), not a rebuilt one
4. Re-derive ids by calling indexOf(item) instead of caching ids across runs

Example fix

// before
E item = index.get(savedId);
// after
int id = index.indexOf(item);
E item2 = (id != -1) ? index.get(id) : null;

if (id < 0 || id >= index.size()) {
  throw new IllegalArgumentException("id " + id + " not present in index of size " + index.size());
}
E item = index.get(id);

// Java has no runtime type guard; use a bounds-check helper
static <E> E safeGet(ConcurrentHashIndex<E> idx, int i) {
  return (i >= 0 && i < idx.size()) ? idx.get(i) : null;
}

try {
  E item = index.get(id);
} catch (ArrayIndexOutOfBoundsException e) {
  log.warn("Stale/unknown id {} against index of size {}", id, index.size());
  item = null; // fall back to re-indexing the item
}

When it happens

Trigger: Calling get(i) with an id returned by indexOf() as UNKNOWN (-1), an id saved to disk for a different index (deserialized via saveToWriter/loadFromReader into a new index), or an id >= size() of the live index.

Common situations: Persisting a model with index ids and loading it against a rebuilt/shorter index; using -1 (UNKNOWN_ID from indexOf) as a lookup key; concurrent modification where indexSize shrank or the caller held an old id.

Related errors

• Couldn't submit item to threadpool:
• Empty index
• Index outside the bounds
• Index not large enough to name all the array elements!
• Invalid phraseColIndex

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/6d0a6f6de696707b. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/util/concurrent/ConcurrentHashIndex.java:69

   */
  public ConcurrentHashIndex(int initialCapacity) {
    item2Index = new ConcurrentHashMap<>(initialCapacity);
    indexSize = 0;
    lock = new ReentrantLock();
    Object[] arr = new Object[initialCapacity];
    index2Item = new AtomicReference<>(arr);
  }

  @SuppressWarnings("unchecked")
  @Override
  public E get(int i) {
    Object[] arr = index2Item.get();
    if (i < indexSize) {
      // arr.length guaranteed to be == to size() given the
      // implementation of indexOf below.
      return (E) arr[i];
    }
    throw new ArrayIndexOutOfBoundsException(String.format("Out of bounds: %d >= %d", i, indexSize));
  }

  @Override
  public int indexOf(E o) {
    Integer id = item2Index.get(o);
    return id == null ? UNKNOWN_ID : id;
  }

  @Override
  public int addToIndex(E o) {
    Integer index = item2Index.get(o);
    if (index != null) {
      return index;
    }

    lock.lock();
    try {
      // Recheck state

View on GitHub (pinned to 1b7edd19c4)