MurmurHash

MurmurHash is a non-cryptographic hash function suitable for general hash-based lookup.^[1][2][3] It was created by Austin Appleby in 2008^[4] and is currently hosted on Github at https://github.com/aappleby/smhasher along with its test suite named 'SMHasher'. It also exists in a number of variants,^[5] all of which have been released into the public domain. The name comes from two basic operations, multiply (MU) and rotate (R), used in its inner loop.

Unlike cryptographic hash functions, it is not specifically designed to be difficult to reverse by an adversary, making it unsuitable for cryptographic purposes.

Variants

The current version is MurmurHash3,^[6][7] which yields a 32-bit or 128-bit hash value.

The older MurmurHash2^[8] yields a 32-bit or 64-bit value. Slower versions of MurmurHash2 are available for big-endian and aligned-only machines. The MurmurHash2A variant adds the Merkle–Damgård construction so that it can be called incrementally. There are two variants which generate 64-bit values; MurmurHash64A, which is optimized for 64-bit processors, and MurmurHash64B, for 32-bit ones. MurmurHash2-160 generates the 160-bit hash, and MurmurHash1 is obsolete.

Implementations

The canonical implementation is in C++, but there are efficient ports for a variety of popular languages, including Python,^[9] C,^[10] Go,^[11] C#,^[7][12] Perl,^[13] Ruby,^[14] Rust,^[15] PHP,^[16] Common Lisp,^[17] Haskell,^[18] Scala,^[19] Java,^[20][21] Erlang,^[22] and JavaScript,^[23][24] together with an online version.^[25]

It has been adopted into a number of open-source projects, most notably libstdc++ (ver 4.6), nginx (ver 1.0.1),^[26] Rubinius,^[27] libmemcached (the C driver for Memcached),^[28] maatkit,^[29] Hadoop,^[1] Kyoto Cabinet,^[30] RaptorDB,^[31] OlegDB,^[32] Cassandra,^[33] Clojure,^[34] Solr,^[35] vowpal wabbit, ^[36]Elasticsearch,^[37] and Guava. ^[38]

Algorithm

Murmur3_32(key, len, seed)
    // Note: In this version, all integer arithmetic is performed with unsigned 32 bit integers.
    //       In the case of overflow, the result is constrained by the application of modulo  arithmetic.
    
    c1 ← 0xcc9e2d51
    c2 ← 0x1b873593
    r1 ← 15
    r2 ← 13
    m ← 5
    n ← 0xe6546b64
 
    hash ← seed

    for each fourByteChunk of key
        k ← fourByteChunk

        k ← k × c1
        k ← (k ROL r1)
        k ← k × c2

        hash ← hash XOR k
        hash ← (hash ROL r2)
        hash ← hash × m + n

    with any remainingBytesInKey
        remainingBytes ← SwapEndianOrderOf(remainingBytesInKey)
        // Note: Endian swapping is only necessary on big-endian machines.
        //       The purpose is to place the meaningful digits towards the low end of the value,
        //       so that these digits have the greatest potential to affect the low range digits
        //       in the subsequent multiplication.  Consider that locating the meaningful digits
        //       in the high range would produce a greater effect upon the high digits of the
        //       multiplication, and notably, that such high digits are likely to be discarded
        //       by the modulo arithmetic under overflow.  We don't want that.
        
        remainingBytes ← remainingBytes × c1
        remainingBytes ← (remainingBytes ROL r1)
        remainingBytes ← remainingBytes × c2

        hash ← hash XOR remainingBytes
 
    hash ← hash XOR len

    hash ← hash XOR (hash >> 16)
    hash ← hash × 0x85ebca6b
    hash ← hash XOR (hash >> 13)
    hash ← hash × 0xc2b2ae35
    hash ← hash XOR (hash >> 16)

A sample C implementation follows

#define ROT32(x, y) ((x << y) | (x >> (32 - y))) // avoid effort
uint32_t murmur3_32(const char *key, uint32_t len, uint32_t seed) {
	static const uint32_t c1 = 0xcc9e2d51;
	static const uint32_t c2 = 0x1b873593;
	static const uint32_t r1 = 15;
	static const uint32_t r2 = 13;
	static const uint32_t m = 5;
	static const uint32_t n = 0xe6546b64;

	uint32_t hash = seed;

	const int nblocks = len / 4;
	const uint32_t *blocks = (const uint32_t *) key;
	int i;
	uint32_t k;
	for (i = 0; i < nblocks; i++) {
		k = blocks[i];
		k *= c1;
		k = ROT32(k, r1);
		k *= c2;

		hash ^= k;
		hash = ROT32(hash, r2) * m + n;
	}

	const uint8_t *tail = (const uint8_t *) (key + nblocks * 4);
	uint32_t k1 = 0;

	switch (len & 3) {
	case 3:
		k1 ^= tail[2] << 16;
	case 2:
		k1 ^= tail[1] << 8;
	case 1:
		k1 ^= tail[0];

		k1 *= c1;
		k1 = ROT32(k1, r1);
		k1 *= c2;
		hash ^= k1;
	}

	hash ^= len;
	hash ^= (hash >> 16);
	hash *= 0x85ebca6b;
	hash ^= (hash >> 13);
	hash *= 0xc2b2ae35;
	hash ^= (hash >> 16);

	return hash;
}

See also

• Fowler–Noll–Vo hash function
• Jenkins hash function
• CityHash

References