Can anyone please check if I am doing this hashing correctly

Question

I am trying to do an implementation of Fowler–Noll–Vo hash function

The Pseudocode looks like this

  hash = FNV_offset_basis
   for each byte_of_data to be hashed
        hash = hash × FNV_prime
        hash = hash XOR byte_of_data
   return hash

This is my code for that

uint8_t            byte_of_data;
uint16_t          hash;
uint16_t          FNV_offset_basis;
uint16_t          FNV_prime;
void computeHash(std::string p)
{
    FNV_offset_basis =  0xcbf29ce484222325;
    FNV_prime        =  0x100000001b3;

    hash = FNV_offset_basis;
    
    //Iterate through the string
    for(int i=0 ; i<p.size();i++)
    {
        hash = hash * FNV_prime;
        hash = hash ^ p.at(i);
    }
   
    std::cout << hash;  //output 2983
     std::cout << std::hex << hash ; //ba7
}

Now I am using it as this

int main()
{
   computeHash("Hello");
}

I am testing my result here and I get the result as 0d47307150c412cf

Update:

I fixed my types to

uint8_t            byte_of_data;
uint64_t          hash;
uint64_t          FNV_offset_basis;
uint64_t          FNV_prime;

and I get the result fa365282a44c0ba7 which still does not match the result 0d47307150c412cf

Any suggestions on how I can fix this

Are there not published test vectors against which you can write unit tests? — Jonathon Reinhart, Sep 18 '15 at 22:55
If your hash is of type `uint16_t`, how do you expect to get a 64 bit long result? — Baum mit Augen, Sep 18 '15 at 22:58
There are options for 32, 64 ... bits. You results comes out when I select 64bit and look at the output of FNV_1. What exactly do you expect to get as output? Mixing the bit length of your data leads to wrong results. :) — rbaleksandar, Sep 18 '15 at 22:58
@JonathonReinhart thanks for pointing that out however I would like to get the correct result printed out first before returning it. — James Franco, Sep 18 '15 at 23:10

deviantfan · Accepted Answer · 2015-09-19T02:16:32.887

Your current result fa365282a44c0ba7 is correct according to the official reference
source code (in C) and manual calculation too... this makes the test site wrong.

The reference source files are linked here: C file and H file
I removed the includes of longlong.h and added following two code parts instead:

/*before the reference code*/

#include <stdint.h>
#define HAVE_64BIT_LONG_LONG
typedef uint64_t u_int64_t;
typedef uint32_t u_int32_t;

/*after it*/

#include<stdio.h>
int main()
{
    printf("%llx\n", fnv_64_str("Hello", FNV1_64_INIT));
}

to make it compile with gcc -std=c11 source.c
(gcc (i686-posix-sjlj-rev0, Built by MinGW-W64 project) 4.9.1)

Output: fa365282a44c0ba7.
And Ideone says so too

score 0 · Answer 2 · answered Sep 18 '15 at 22:58

0

This is the problem:

uint16_t          FNV_offset_basis;
uint16_t          FNV_prime;
void computeHash(std::string p)
{
    FNV_offset_basis =  0xcbf29ce484222325;
    FNV_prime        =  0x100000001b3;

FNV_prime and FNV_offset_basis are both 16-bit integers in your code, yet inexplicably you're assigning long 64-bit integers to them, your C++ compiler should be warning you about an improper literal assignment.

What happens if your change the types to uint64_t?

answered Sep 18 '15 at 22:58

Dai

141,631
28
261
374

And the hash variable too, not just the two vars shown here. – deviantfan Sep 18 '15 at 22:59
Yeah, it all needs to be 64bit if 64bit is intended to be used. – rbaleksandar Sep 18 '15 at 22:59
So I change the types to `uint8_t byte_of_data; uint64_t hash; uint64_t FNV_offset_basis; uint64_t FNV_prime;` Now I get the result fa365282a44c0ba7 which is still incorrect – James Franco Sep 18 '15 at 23:05

Can anyone please check if I am doing this hashing correctly

2 Answers2