将C哈希函数转换为PHP：ezmlm_hash

Question

I'm trying to upgrade from PHP 5.2.x to 5.3.2 on my server. Problem is, I relying on the broken implementation of PHP's ezmlm_hash() (the bug is outlined here: http://bugs.php.net/bug.php?id=47969).

My first thought was to rewrite the broken version of the native PHP function (which is written in C) myself in PHP and use that in my code, instead of modifying the PHP source code and having to compile PHP from source.

Here is the C version of the code:

PHP_FUNCTION(ezmlm_hash)
{
    char *str = NULL;
    unsigned int h = 5381L;
    int j, str_len;

    if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s",
                              &str, &str_len) == FAILURE) {
        return;
    }

    for (j = 0; j < str_len; j++) {
        h = (h + (h << 5)) ^ (unsigned long) (unsigned char) tolower(str[j]);
    }

    h = (h % 53);

    RETURN_LONG((int) h);
}

here is what I've written in PHP:

function ezmlm_hash_mine($email_address){
    $h = 5381;
    $email_length = strlen($email_address);
    for($x=0;$x<$email_length;$x++){
        $chr = strtolower($email_address[$x]);
        $h = ($h + ($h << 5)) ^ ( ord($chr) );
    }

    $h = $h % 53;
    return $h;
}

I'm using a 64-bit machine. The two functions output different results:

$email_addresses = array(
    '[e-mail]',
    '[e-mail]',
);

print('<PRE>');

foreach($email_addresses as $email_address){
    print(ezmlm_hash($email_address).PHP_EOL);
    print(ezmlm_hash_mine($email_address).PHP_EOL.PHP_EOL);
}

output:

23
-52

15
-21

I know I probably have some precision or typing issues, I'm just not sure how to fix it. Any help would be greatly appreciated!

UPDATE

When I run thes the code on 32 bit machines, they both output the new corrected values:

12
12

45
45

I think this has something to do with the modulo operator... does anyone know the PHP equivalent of the C modulo operator? % in PHP behaves differently!

UPDATE 2

It appears as if this is not possible with vanilla PHP, as it's floating point arithmetic doesn't have enough precision, and weirdness in . I'll have to install either BCMath or GMP. Thanks for everyone's insight.

Answer 1

try this EDIT truncate to 32 bits after calculation:

function ezmlm_hash_mine($email_address){
    $h = gmp_init(5381);
    $d = gmp_setbit(0, 64);
    $d32 = gmp_setbit(0, 32);
    $email_length = strlen($email_address);

    $chr = strtolower($email_address);

    for($x=0;$x<$email_length;$x++){    
        $h = gmp_mod(gmp_xor(gmp_mod(gmp_add($h, gmp_mod(gmp_mul($h, "32"), $d)), $d), ord($chr[$x])), $d32);
    }

    $h = gmp_mod($h, 53);
    return gmp_intval($h);
}