What kind of hashing function/algorithm does javascript associative array use?

Question

We learned there are many different hashing algorithms/functions, I'm curious which one is used by javascript (v8, if the implementation matters).

Answer 1

Since V8 is open source, you go to the source:

Here's GetHash(): https://github.com/v8/v8/blob/master/src/objects.cc#L903

And, here's are some of the hash functions for different types: https://github.com/v8/v8-git-mirror/blob/bda7fb22465fc36d99b4053f0ef60cfaa8441209/src/utils.h#L347

And, this looks like the core hash computation for strings: https://code.google.com/p/v8/source/browse/trunk/src/objects.cc?spec=svn6&r=6#3574

uint32_t String::ComputeHashCode(unibrow::CharacterStream* buffer,
                                 int length) {
  // Large string (please note large strings cannot be an array index).
  if (length > kMaxMediumStringSize) return HashField(length, false);

  // Note: the Jenkins one-at-a-time hash function
  uint32_t hash = 0;
  while (buffer->has_more()) {
    uc32 r = buffer->GetNext();
    hash += r;
    hash += (hash << 10);
    hash ^= (hash >> 6);
  }
  hash += (hash << 3);
  hash ^= (hash >> 11);
  hash += (hash << 15);

  // Short string.
  if (length <= kMaxShortStringSize) {
    // Make hash value consistent with value returned from String::Hash.
    buffer->Rewind();
    uint32_t index;
    hash = HashField(hash, ComputeArrayIndex(buffer, &index, length));
    hash = (hash & 0x00FFFFFF) | (length << kShortLengthShift);
    return hash;
  }

  // Medium string (please note medium strings cannot be an array index).
  ASSERT(length <= kMaxMediumStringSize);
  // Make hash value consistent with value returned from String::Hash.
  hash = HashField(hash, false);
  hash = (hash & 0x0000FFFF) | (length << kMediumLengthShift);
  return hash;
}

It's probably worth mentioning that V8 goes to great lengths to avoid using a hash table for object properties whenever possible, preferring to compile known property references into direct index references rather than run-time hash lookups for performance reasons (though this is only possible sometimes - depends upon the code).