This code implements MurmurHash3's 32-bit hashing function in pure bash.

For short strings, bash can be 10 times faster than a C implementation, because there is no executable overhead.

As string length increases, the bash implementation exhibits a quadratic increase in runtime. This is because indexing strings in bash is linear in the size of the index.

Benchmarks (seconds, lower is better; * = best):

                                bash-utf8    bash-ascii       C
  20000 hashes, ASCII words        1.86         1.58 *      19.17
1000 hashes, 100 bytes UTF8        1.01         0.36 *       0.91
         1 hash,  1KB  UTF8        0.020        0.009        0.002 *        
         1 hash, 10KB  UTF8        1.63         0.77         0.005 *
         1 hash, 20KB  UTF8        6.31         3.02         0.010 *

Usage

source mumu3-32-utf8.sh

source test.sh # optional!

mumu "€𠳏b✗a" # prints da7915bd
echo "$hash"  # hash=3665368509, 0xda7915bd in decimal

mumu "€𠳏b✗a" 0xABCDEF      # with initial seed
mumu "€𠳏b✗a" "" ""         # no initial seed, suppress output, output in $hash only
mumu "€𠳏b✗a" "" "%d\n"     # prints 3665368509: non-empty third argument is a printf format
mumu "€𠳏b✗a" "" "0x%08X\n" # prints 0xDA7915BD

Notes

• test.sh needs a reference implementation (recommend Peter Scott's MurmurHash3 in C) to perform random tests. If absent, only the hardcoded tests are performed.
• the ASCII-only version is much shorter, but only slightly faster for ASCII-only input, because the UTF8 version will fall back to the faster ASCII-only algorithm if the input allows it. ASCII function can still hash UTF8 inputs, but gives different results than a reference implementation, due to byte ordering issues.