Hex-Pair Hamming Distance Functions

This extension provides four functions to work with hex representation of feature vector.

Every function asserts that both arguments have the same length.

hamming_distance(varchar, varchar) RETURNS int (IMMUTABLE STRICT)

Computes absolute hamming distance between to FVs

SELECT hamming_distance('1d3f', '1110');
 hamming_distance
------------------
                7
(1 row)

This expands as

a = '1d3f'  -> 0001 1101 0011 1111
b = '1110'  -> 0001 0001 0001 0000

c = a ^ b   -> 0000 1100 0010 1111

popcount(c) ->   0 + 2 + 1 + 4 = 7

hamming_similarity(varchar, varchar) RETURNS int (IMMUTABLE STRICT)

SELECT hamming_similarity('1d3f', '1110');
 hamming_similarity
--------------------
                  9
(1 row)

Can be defined as

HAMMING_SIMILARITY = len(a) * 4 - hamming_distance(a, b);

hamming_distance_normalized(varchar, varchar) RETURNS float (IMMUTABLE STRICT)

SELECT hamming_distance_normalized('1d3f', '1110');
 hamming_distance_normalized
-----------------------------
                      0.4375
(1 row)

Can be defined as

HAMMING_DISTANCE_NORM = hamming_distance(a, b) / len(a) * 4;

hamming_similarity_normalized(varchar, varchar) RETURNS float (IMMUTABLE STRICT)

SELECT hamming_similarity_normalized('1d3f', '1110');
 hamming_similarity_normalized
-------------------------------
                        0.5625
(1 row)

Can be defined as

HAMMING_SIMILARITY_NORM = hamming_similarity(a, b) / len(a) * 4;

Installation

cd .../hamming_distance
sudo make install

Running tests

make installcheck