Rolling hash

A rolling hash is a hash function where the input is hashed in a window that moves through the input.

A few hash functions allow a rolling hash to be computed very quickly -- the new hash value is rapidly calculated given only the old hash value, the old value removed from the window, and the new value added to the window -- similar to the way a moving average function can be computed much more quickly than other low-pass filters.

One of the main applications is the Rabin-Karp string search algorithm, which uses the rolling hash described below.

Another popular application is rsync program which uses a checksum based on Mark Adler's adler-32 as its rolling hash.

Contents 1 Rabin-Karp rolling hash 2 Cyclic polynomial 3 Independence 4 Computational complexity 5 Software 6 Footnotes

[edit] Rabin-Karp rolling hash

The Rabin-Karp string search algorithm is normally used with a very simple rolling hash function that only uses multiplications and additions:

H = c₁a^{k − 1} + c₂a^{k − 2} + c₃a^{k − 3} + ... + c_ka⁰ where a is a constant and c₁,...,c_k are the input characters.

In order to avoid manipulating huge H values, all math is done modulo n. The choice of a and n is critical to get good hashing; see linear congruential generator for more discussion.

Removing and adding chars simply involves adding or subtracting the first or last term. Shifting all chars by one position to the left requires multiplying the entire sum H by a. Shifting all chars by one position to the right requires dividing the entire sum H by a. Note that in modulo arithmetic, a can be chosen to have a multiplicative inverse a ^{− 1} by which H can be multiplied to get the result of the division without actually performing a division.

[edit] Cyclic polynomial

Hashing by cyclic polynomial^[1]—sometimes called Buzhash—is also simple, but it has the benefit of avoiding multiplications, using barrel shifts instead. It presumes that there is some hash function h from characters to integers in the interval [0,2^L). This hash function might be simply an array or a hash table mapping characters to random integers. Let the function s be a cyclic binary rotation (or barrel shift): it rotates the bits by 1 to the left, pushing the latest bit in the first position. E.g., s(10011) = 00111. Let be the bit-wise exclusive or. The hash values are defined as

where the multiplications by powers of two can be implemented by binary shifts. The result is a number in [0,2^L).

Computing the hash values in a rolling fashion is done as follows. Let H be the previous hash value. Rotate H once: . If c₁ is the character to be removed, rotate it k times: s^k(h(c₁)). Then simply set

where c_{k + 1} is the new character.

[edit] Independence

At best, rolling hash values are pairwise independent^[2]. Similarly, at best the (randomized) Rabin-Karp rolling hash values are uniform. Hashing n-grams by cyclic polynomials achieves pairwise independence: simply keep the first L − n + 1 bits.

[edit] Computational complexity

All rolling hash functions are linear in the number of characters, but their complexity with respect to the length of the window (k) varies. Rabin-Karp rolling hash requires the multiplications of two k-bit numbers, integer multiplication is in ^[3]. Hashing n-grams by cyclic polynomials can be done in linear time ^[2].

[edit] Software

• ngramhashing is a Free software C++ implementation of several rolling hash functions

[edit] Footnotes

1. ^ Jonathan D. Cohen, Recursive hashing functions for n-grams, ACM Trans. Inf. Syst. 15 (3), 1997
2. ^ ^{a b} Daniel Lemire, Owen Kaser: Recursive n-gram hashing is pairwise independent, at best, Computer Speech & Language 24 (4), pages 698-710, 2010. arXiv:0705.4676
3. ^ M. Fürer, Faster integer multiplication, in: STOC ’07, 2007, pp. 57–66.