文字

uniqid

(PHP 4, PHP 5)

uniqid生成一个唯一ID

说明

string uniqid ([ string $prefix = "" [, bool $more_entropy = false ]] )

获取一个带前缀、基于当前时间微秒数的唯一ID。

参数

prefix

有用的参数。例如:如果在多台主机上可能在同一微秒生成唯一ID。

prefix为空,则返回的字符串长度为13。more_entropy TRUE ,则返回的字符串长度为23。

more_entropy

如果设置为 TRUE uniqid() 会在返回的字符串结尾增加额外的煽(使用combined linear congruential generator)。 使得唯一ID更具唯一性。

返回值

返回字符串形式的唯一ID。

范例

Example #1 uniqid() 例子

<?php

printf ( "uniqid(): %s\r\n" uniqid ());


printf ( "uniqid('php_'): %s\r\n" uniqid ( 'php_' ));


printf ( "uniqid('', true): %s\r\n" uniqid ( '' true ));
?>

更新日志

版本 说明
5.0.0 prefix 参数设为可选。
4.3.1 prefix 参数的长度限制提升到114个字符。

注释

Note:

在Cygwin环境下,为了使此函数能够工作,more_entropy 必须设置为 TRUE

用户评论:

[#1] jensenv at cshs dot org [2015-05-01 00:26:05]

DUPLICATE UID WARNING: Be aware that calling this function many times in a tight loop (for instance, to assign UIDs to objects in an array) can result in many of the UIDs being identical, since less than a microsecond may have passed since the previous call. One simple solution is to compare the latest value returned with the previous result, and keep calling in a loop until the new value is different, if you don't mind this call causing a delay of 1 microsecond per call when necessary.

[#2] pe_ank at peanksoft dot com (Hengky Gumilar) [2014-03-13 05:43:12]

I guess, this is not a bugs, just a note for someone newbie like me, 
i've got problem on uploading image, i tested on local server, it works fine but when i upload on my hosting site , the scripts failed (i use shared hosting, so i' m helpless), this is a part of my scripts:

$ext=$this->_fuphandler->extractExtension($data);
$strsha=sha1($this->_fuphandler->extractFilename($data));
$r=sprintf($pathToRoot.$dest.'%s.%s',$strsha,$ext); // filename builder
$rawsrc = $data['tmp_name'];
if(!move_uploaded_file($rawsrc,$r))
{....}
it should work,  it did on my uniform server, but it didn't on my hosting so i spend many hours on trying, then i try  to change the  $strsha=uniqid();
and it works...
i know it's just a silly thing, ...hehehe, beside i'm just a newbie, but probably will help someone who has some problem like me.

[#3] team at researchbib dot com [2014-01-10 00:53:44]

The following is one example to get a random string class.


class RandomString
{

    
    static public function getRandom($length)
    {
        $randomData = base64_encode(
                file_get_contents('/dev/urandom', false, null, 0, $length) . uniqid(mt_rand(), true)
        );
        return substr(strtr($randomData, array('/' => '', '=' => '', '+' => '')), 0, $length);
    }

}

[#4] ciantic at NOSPAM dot oksidi dot com [2013-12-05 15:34:19]

If you are storing the value to database it might be more efficient to use more letters than in hexadecimal, for me I decided 0-9a-z is good enough.

Notice also that many databases are case insensitive by default so using capitals maybe unwise anyway:

<?php

function uniqid_base36($more_entropy=false) {
    
$s uniqid(''$more_entropy);
    if (!
$more_entropy)
        return 
base_convert($s1636);
        
    
$hex substr($s013);
    
$dec $s[13] . substr($s15); // skip the dot
    
return base_convert($hex1636) . base_convert($dec1036);
}

echo 
uniqid_base36() . "\n"// ?eb98xzzhq7
echo uniqid_base36(true) . "\n"// eb98xzzhr8dervre

?>


String length may vary with this method.

[#5] Anonymous [2012-11-22 18:52:40]

I wouldn't use:
$prefix = exec("hostname");
- since it's CPU-expensive.

Instead use php_uname('n') if you have many visitors.

[#6] Anonymous [2012-08-16 14:26:25]

If you're on a Windows system, and you wan't the prefix to be the computer name (and not only localhost), you can invoke the exec function to get it.

Then your unique id should be :

<?php
$prefix 
exec("hostname");
echo 
uniqid($prefixtrue);
?>


Regards

[#7] redtraider at gmail dot com [2012-02-14 12:37:56]

I use this function to generate microsoft-compatible GUID's.

<?php
 
public function create_guid($namespace '') {     
    static 
$guid '';
    
$uid uniqid(""true);
    
$data $namespace;
    
$data .= $_SERVER['REQUEST_TIME'];
    
$data .= $_SERVER['HTTP_USER_AGENT'];
    
$data .= $_SERVER['LOCAL_ADDR'];
    
$data .= $_SERVER['LOCAL_PORT'];
    
$data .= $_SERVER['REMOTE_ADDR'];
    
$data .= $_SERVER['REMOTE_PORT'];
    
$hash strtoupper(hash('ripemd128'$uid $guid md5($data)));
    
$guid '{' .   
            
substr($hash,  0,  8) . 
            
'-' .
            
substr($hash,  8,  4) .
            
'-' .
            
substr($hash12,  4) .
            
'-' .
            
substr($hash16,  4) .
            
'-' .
            
substr($hash2012) .
            
'}';
    return 
$guid;
  }
?>

[#8] tom [2011-06-17 12:40:03]

mt_rand() features in a lot of comments here. It is a good PRNG for Monte Carlo simulations, not for anything related to security. Wikipedia's page on cryptographically-secure PRNGs explains. If you want it to be hard for an attacker to guess or predict a "random" UUID, try using /dev/random instead.

<?php
$r 
unpack('v*'fread(fopen('/dev/random''r'),16));
$uuid sprintf('%04x%04x-%04x-%04x-%04x-%04x%04x%04x',
    
$r[1], $r[2], $r[3], $r[4] & 0x0fff 0x4000
    
$r[5] & 0x3fff 0x8000$r[6], $r[7], $r[8]);
?>


That's obviously not production code. It's just for illustration.

[#9] Anonymous [2011-06-16 16:08:27]

Prefix can be useful, for instance, if you generate identifiers simultaneously on several hosts that might happen to generate the identifier at the same microsecond.
So we can include the hostname / servername in the id.
<?php
echo uniqid(php_uname('n'), true);
// Output: darkstar4dfa8c27aea106.40781203
?>

[#10] rommel at rommelsantor dot com [2011-05-28 15:29:27]

The php5-uuid functions could definitely use some documentation to clarify how they should be used, but here's what I've gleaned by examining the OSSP source code (found here: http://ossp-uuid.sourcearchive.com/documentation/1.5.1-1ubuntu1/php_2uuid_8c-source.html).

The uuid_make() function takes two arguments when generating v1 or v4, but four arguments are required when generating v3 or v5. The first two arguments have been demonstrated below and are straightforward, so I'll skip to the as-yet non-described arguments.

The third argument to uuid_make() is: $namespace
  - this is a secondary resource created with uuid_create(); it is apparently used to generate an internal UUID, which is used as the namespace of the output UUID

The fourth argument to uuid_make() is: $url
  - this is the value that is to be hashed (MD5 for v3, SHA-1 for v5); it may be any string or even null

Here's a simple class illustrating the proper usage (note that if php5-uuid is not installed on your system, each function call will just return false):

<?php
class UUID {
  

  
public static function v1() {
    if (!
function_exists('uuid_create'))
      return 
false;

    
uuid_create(&$context);
    
uuid_make($contextUUID_MAKE_V1);
    
uuid_export($contextUUID_FMT_STR, &$uuid);
    return 
trim($uuid);
  }

  

  
public static function v3($i_url) {
    if (!
function_exists('uuid_create'))
      return 
false;

    if (!
strlen($i_url))
      
$i_url self::v1();

    
uuid_create(&$context);
    
uuid_create(&$namespace);

    
uuid_make($contextUUID_MAKE_V3$namespace$i_url);
    
uuid_export($contextUUID_FMT_STR, &$uuid);
    return 
trim($uuid);
  }

  

  
public static function v4() {
    if (!
function_exists('uuid_create'))
      return 
false;

    
uuid_create(&$context);

    
uuid_make($contextUUID_MAKE_V4);
    
uuid_export($contextUUID_FMT_STR, &$uuid);
    return 
trim($uuid);
  }

  

  
public static function v5($i_url) {
    if (!
function_exists('uuid_create'))
      return 
false;

    if (!
strlen($i_url))
      
$i_url self::v1();

    
uuid_create(&$context);
    
uuid_create(&$namespace);

    
uuid_make($contextUUID_MAKE_V5$namespace$i_url);
    
uuid_export($contextUUID_FMT_STR, &$uuid);
    return 
trim($uuid);
  }
}
?>


And here's a demonstration:

<?php
for ($i 1$i <= 3; ++$i) {
  echo 
'microtime = ' microtime(true) . '<br/>';
  echo 
"V1 UUID: " UUID::v1() . '<br/>';
  echo 
"V3 UUID of URL='abc': " UUID::v3('abc') . '<br/>';
  echo 
"V4 UUID: " UUID::v4() . '<br/>';
  echo 
"V5 UUID of URL=null: " UUID::v5(null) . '<br/>';
  echo 
'<hr/>';
}
?>


And the output:

microtime = 1306620716.0457
V1 UUID: 7fddae8e-8977-11e0-bc11-003048c3b1f2
V3 UUID of URL='abc': 522ec739-ca63-3ec5-b082-08ce08ad65e2
V4 UUID: b3851ec7-4871-4527-92b5-ef5616bae1e6
V5 UUID of URL=null: e129f27c-5103-5c5c-844b-cdf0a15e160d
-------------------
microtime = 1306620716.0465
V1 UUID: 7fddb83e-8977-11e0-9e6e-003048c3b1f2
V3 UUID of URL='abc': 522ec739-ca63-3ec5-b082-08ce08ad65e2
V4 UUID: 7e78fe0d-59b8-4637-af7f-e88d221a7d1e
V5 UUID of URL=null: e129f27c-5103-5c5c-844b-cdf0a15e160d
-------------------
microtime = 1306620716.0467
V1 UUID: 7fddbfb4-8977-11e0-a2bc-003048c3b1f2
V3 UUID of URL='abc': 522ec739-ca63-3ec5-b082-08ce08ad65e2
V4 UUID: 12a940c7-0f3f-46a1-bb5f-bdd602e10654
V5 UUID of URL=null: e129f27c-5103-5c5c-844b-cdf0a15e160d

As you can see, the calls to v3() always return the same UUID because the same URL parameter, "abc", is always supplied. The same goes for the v5() function which is always supplied a null URL.

The v4() UUIDs are always entirely different because they are (pseudo)random. And the v1() calls are very similar but just slightly different because it's based on the computer's MAC address and the current time.

[#11] php at ryanmckeel dot com [2010-06-29 10:15:17]

Calls to uuid_make that use the constants UUID_MAKE_V5 or UUID_MAKE_V3 (using Debian package php5-uuid available June 2010) will not work with only two variables.

I could not find good documentation, so I read some source code and figured out that this would work:

uuid_create(&$v5);
//uuid_make($v5, UUID_MAKE_V5);
uuid_make($v5, UUID_MAKE_V5,$v5,$uniqid());
uuid_export($v5, UUID_FMT_STR, &$v5String);

Please use at your own risk.  This may not be the best way to give this param variables, but it at least makes it work what appears to be properly (generating unique ID's).

I imagine that UUID_MAKE_V3 is similar in what it needs.

  Ryan

[#12] Enrico Pallazzo [2010-03-21 14:45:33]

Running:

rand_uniqid(9007199254740989);

will return 'PpQXn7COf' and:

rand_uniqid('PpQXn7COf', true);

will return '9007199254740989'

---

If you want the rand_uniqid to be at least 6 letter long, use the $pad_up = 6 argument

---

You can support even more characters (making the resulting rand_uniqid even smaller) by adding characters to the $index var at the top of the function body.

---

<?php
function rand_uniqid($in$to_num false$pad_up false$passKey null)
{
    
$index "abcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
    if (
$passKey !== null) {
        
// Although this function's purpose is to just make the
        // ID short - and not so much secure,
        // you can optionally supply a password to make it harder
        // to calculate the corresponding numeric ID

        
for ($n 0$n<strlen($index); $n++) {
            
$i[] = substr$index,$n ,1);
        }

        
$passhash hash('sha256',$passKey);
        
$passhash = (strlen($passhash) < strlen($index))
            ? 
hash('sha512',$passKey)
            : 
$passhash;

        for (
$n=0$n strlen($index); $n++) {
            
$p[] =  substr($passhash$n ,1);
        }

        
array_multisort($p,  SORT_DESC$i);
        
$index implode($i);
    }

    
$base  strlen($index);

    if (
$to_num) {
        
// Digital number  <<--  alphabet letter code
        
$in  strrev($in);
        
$out 0;
        
$len strlen($in) - 1;
        for (
$t 0$t <= $len$t++) {
            
$bcpow bcpow($base$len $t);
            
$out   $out strpos($indexsubstr($in$t1)) * $bcpow;
        }

        if (
is_numeric($pad_up)) {
            
$pad_up--;
            if (
$pad_up 0) {
                
$out -= pow($base$pad_up);
            }
        }
        
$out sprintf('%F'$out);
        
$out substr($out0strpos($out'.'));
    } else {
        
// Digital number  -->>  alphabet letter code
        
if (is_numeric($pad_up)) {
            
$pad_up--;
            if (
$pad_up 0) {
                
$in += pow($base$pad_up);
            }
        }

        
$out "";
        for (
$t floor(log($in$base)); $t >= 0$t--) {
            
$bcp bcpow($base$t);
            
$a   floor($in $bcp) % $base;
            
$out $out substr($index$a1);
            
$in  $in - ($a $bcp);
        }
        
$out strrev($out); // reverse
    
}

    return 
$out;
}

echo 
rand_uniqid(1);
?>

[#13] keith at keithtyler dot com [2009-12-07 18:18:57]

For the record, the underlying function to uniqid() appears to be roughly as follows:

$m=microtime(true);
sprintf("%8x%05x\n",floor($m),($m-floor($m))*1000000);

In other words, first 8 hex chars = Unixtime, last 5 hex chars = microseconds. This is why it has microsecond precision. Also, it provides a means by which to reverse-engineer the time when a uniqid was generated: 

date("r",hexdec(substr(uniqid(),0,8)));

Increasingly as you go further down the string, the number becomes "more unique" over time, with the exception of digit 9, where numeral prevalence is 0..3>4>5..f, because of the difference between 10^6 and 16^5 (this is presumably true for the remaining digits as well but much less noticeable).

[#14] Andrew Moore [2009-12-04 08:45:11]

The following class generates VALID RFC 4211 COMPLIANT Universally Unique IDentifiers (UUID) version 3, 4 and 5.

Version 3 and 5 UUIDs are named based. They require a namespace (another valid UUID) and a value (the name). Given the same namespace and name, the output is always the same.

Version 4 UUIDs are pseudo-random.

UUIDs generated below validates using OSSP UUID Tool, and output for named-based UUIDs are exactly the same. This is a pure PHP implementation.

<?php
class UUID {
  public static function 
v3($namespace$name) {
    if(!
self::is_valid($namespace)) return false;

    
// Get hexadecimal components of namespace
    
$nhex str_replace(array('-','{','}'), ''$namespace);

    
// Binary Value
    
$nstr '';

    
// Convert Namespace UUID to bits
    
for($i 0$i strlen($nhex); $i+=2) {
      
$nstr .= chr(hexdec($nhex[$i].$nhex[$i+1]));
    }

    
// Calculate hash value
    
$hash md5($nstr $name);

    return 
sprintf('%08s-%04s-%04x-%04x-%12s',

      
// 32 bits for "time_low"
      
substr($hash08),

      
// 16 bits for "time_mid"
      
substr($hash84),

      
// 16 bits for "time_hi_and_version",
      // four most significant bits holds version number 3
      
(hexdec(substr($hash124)) & 0x0fff) | 0x3000,

      
// 16 bits, 8 bits for "clk_seq_hi_res",
      // 8 bits for "clk_seq_low",
      // two most significant bits holds zero and one for variant DCE1.1
      
(hexdec(substr($hash164)) & 0x3fff) | 0x8000,

      
// 48 bits for "node"
      
substr($hash2012)
    );
  }

  public static function 
v4() {
    return 
sprintf('%04x%04x-%04x-%04x-%04x-%04x%04x%04x',

      
// 32 bits for "time_low"
      
mt_rand(00xffff), mt_rand(00xffff),

      
// 16 bits for "time_mid"
      
mt_rand(00xffff),

      
// 16 bits for "time_hi_and_version",
      // four most significant bits holds version number 4
      
mt_rand(00x0fff) | 0x4000,

      
// 16 bits, 8 bits for "clk_seq_hi_res",
      // 8 bits for "clk_seq_low",
      // two most significant bits holds zero and one for variant DCE1.1
      
mt_rand(00x3fff) | 0x8000,

      
// 48 bits for "node"
      
mt_rand(00xffff), mt_rand(00xffff), mt_rand(00xffff)
    );
  }

  public static function 
v5($namespace$name) {
    if(!
self::is_valid($namespace)) return false;

    
// Get hexadecimal components of namespace
    
$nhex str_replace(array('-','{','}'), ''$namespace);

    
// Binary Value
    
$nstr '';

    
// Convert Namespace UUID to bits
    
for($i 0$i strlen($nhex); $i+=2) {
      
$nstr .= chr(hexdec($nhex[$i].$nhex[$i+1]));
    }

    
// Calculate hash value
    
$hash sha1($nstr $name);

    return 
sprintf('%08s-%04s-%04x-%04x-%12s',

      
// 32 bits for "time_low"
      
substr($hash08),

      
// 16 bits for "time_mid"
      
substr($hash84),

      
// 16 bits for "time_hi_and_version",
      // four most significant bits holds version number 5
      
(hexdec(substr($hash124)) & 0x0fff) | 0x5000,

      
// 16 bits, 8 bits for "clk_seq_hi_res",
      // 8 bits for "clk_seq_low",
      // two most significant bits holds zero and one for variant DCE1.1
      
(hexdec(substr($hash164)) & 0x3fff) | 0x8000,

      
// 48 bits for "node"
      
substr($hash2012)
    );
  }

  public static function 
is_valid($uuid) {
    return 
preg_match('/^\{?[0-9a-f]{8}\-?[0-9a-f]{4}\-?[0-9a-f]{4}\-?'.
                      
'[0-9a-f]{4}\-?[0-9a-f]{12}\}?$/i'$uuid) === 1;
  }
}

// Usage
// Named-based UUID.

$v3uuid UUID::v3('1546058f-5a25-4334-85ae-e68f2a44bbaf''SomeRandomString');
$v5uuid UUID::v5('1546058f-5a25-4334-85ae-e68f2a44bbaf''SomeRandomString');

// Pseudo-random UUID

$v4uuid UUID::v4();
?>

[#15] John Haugeland from FullOf.BS [2009-05-26 10:11:15]

Generating an MD5 from a unique ID is naive and reduces much of the value of unique IDs, as well as providing significant (attackable) stricture on the MD5 domain.  That's a deeply broken thing to do.  The correct approach is to use the unique ID on its own; it's already geared for non-collision.

IDs should never be obfuscated for security, so if you're worried about someone guessing your ID, fix the system, don't just make it harder to guess (because it's nowhere near as difficult to guess as you imagine: you can just brute force the 60,000 MD5s that are generatable from millisecond IDs over the course of a given minute, which the typical computer can do in less than 0.1s).

If you absolutely need to involve a hash somehow - maybe to placate a boss who thinks they understand security much better than they actually do - append it instead.

function BadIdeaID() { return uniqid() . '_' . md5(mt_rand()); }

[#16] Marius Karthaus [2009-01-23 11:10:47]

Better yet, just use the php5-uuid extension and this class to create 'official' UUIDs at high speed (On my system 1000 UUIDs are created in 0.0064 seconds)

<?php

Usage
:
$uuid=new uuid();
echo 
$uuid->v1();

class 
uuid {
    

    
    
protected $uuidobject;
    
    

    
protected function create() {
        if (! 
is_resource $this->uuidobject )) {
            
uuid_create ( &$this->uuidobject );
        }
    }
    
    

    
public function v1() {
        
$this->create ();
        
uuid_make $this->uuidobjectUUID_MAKE_V1 );
        
uuid_export $this->uuidobjectUUID_FMT_STR, &$uuidstring );
        return 
trim $uuidstring );
    }
    
    

    
public function v4() {
        
$this->create ();
        
uuid_make $this->uuidobjectUUID_MAKE_V4 );
        
uuid_export $this->uuidobjectUUID_FMT_STR, &$uuidstring );
        return 
trim $uuidstring );
    }
    
    

    
public function v5() {
        
$this->create ();
        
uuid_make $this->uuidobjectUUID_MAKE_V5 );
        
uuid_export $this->uuidobjectUUID_FMT_STR, &$uuidstring );
        return 
trim $uuidstring );
    }
}
?>

[#17] Marius Karthaus [2009-01-22 05:39:55]

I've created a class from the function below. With it, creating a lot of uuids beomes about 100 times faster because you do not need to fopen() for each uuid. 
uuid::get() still works as a static function if you just need a single uuid.

<?php
class uuid {

protected $urand;

public function __construct() {
$this->urand = @fopen ( '/dev/urandom', 'rb' );
}


function get() {

$pr_bits = false;
if (is_a ( $this, 'uuid' )) {
if (is_resource ( $this->urand )) {
$pr_bits .= @fread ( $this->urand, 16 );
}
}
if (! $pr_bits) {
$fp = @fopen ( '/dev/urandom', 'rb' );
if ($fp !== false) {
$pr_bits .= @fread ( $fp, 16 );
@fclose ( $fp );
} else {
// If /dev/urandom isn't available (eg: in non-unix systems), use mt_rand().
$pr_bits = "";
for($cnt = 0; $cnt < 16; $cnt ++) {
$pr_bits .= chr ( mt_rand ( 0, 255 ) );
}
}
}
$time_low = bin2hex ( substr ( $pr_bits, 0, 4 ) );
$time_mid = bin2hex ( substr ( $pr_bits, 4, 2 ) );
$time_hi_and_version = bin2hex ( substr ( $pr_bits, 6, 2 ) );
$clock_seq_hi_and_reserved = bin2hex ( substr ( $pr_bits, 8, 2 ) );
$node = bin2hex ( substr ( $pr_bits, 10, 6 ) );


$time_hi_and_version = hexdec ( $time_hi_and_version );
$time_hi_and_version = $time_hi_and_version >> 4;
$time_hi_and_version = $time_hi_and_version | 0x4000;


$clock_seq_hi_and_reserved = hexdec ( $clock_seq_hi_and_reserved );
$clock_seq_hi_and_reserved = $clock_seq_hi_and_reserved >> 2;
$clock_seq_hi_and_reserved = $clock_seq_hi_and_reserved | 0x8000;

return sprintf ( '%08s-%04s-%04x-%04x-%012s', $time_low, $time_mid, $time_hi_and_version, $clock_seq_hi_and_reserved, $node );
}

}

[#18] sean at seancolombo dot com [2009-01-06 07:56:10]

This builds slightly on david's post below.  The differences are that it doesn't require Cake anymore and there is a graceful fallback for /dev/urandom in case that isn't available (/dev/urandom is not available on windows systems for example).

Since it uses mt_rand(), it is still cryptographically secure.

<?php

      function uuidSecure() {
       
        $pr_bits = null;
        $fp = @fopen('/dev/urandom','rb');
        if ($fp !== false) {
            $pr_bits .= @fread($fp, 16);
            @fclose($fp);
        } else {
            // If /dev/urandom isn't available (eg: in non-unix systems), use mt_rand().
            $pr_bits = "";
            for($cnt=0; $cnt < 16; $cnt++){
                $pr_bits .= chr(mt_rand(0, 255));
            }
        }
       
        $time_low = bin2hex(substr($pr_bits,0, 4));
        $time_mid = bin2hex(substr($pr_bits,4, 2));
        $time_hi_and_version = bin2hex(substr($pr_bits,6, 2));
        $clock_seq_hi_and_reserved = bin2hex(substr($pr_bits,8, 2));
        $node = bin2hex(substr($pr_bits,10, 6));
       
        
        $time_hi_and_version = hexdec($time_hi_and_version);
        $time_hi_and_version = $time_hi_and_version >> 4;
        $time_hi_and_version = $time_hi_and_version | 0x4000;
       
        
        $clock_seq_hi_and_reserved = hexdec($clock_seq_hi_and_reserved);
        $clock_seq_hi_and_reserved = $clock_seq_hi_and_reserved >> 2;
        $clock_seq_hi_and_reserved = $clock_seq_hi_and_reserved | 0x8000;
       
        return sprintf('%08s-%04s-%04x-%04x-%012s',
            $time_low, $time_mid, $time_hi_and_version, $clock_seq_hi_and_reserved, $node);
    }

[#19] david at ramaboo dot com [2009-01-05 00:51:33]

Another UUID function. This time using /dev/random
<?php

    
public function uuidSecure() {
        
        
$pr_bits null;
        
$fp = @fopen('/dev/urandom','rb');
        if (
$fp !== false) {
            
$pr_bits .= @fread($fp16);
            @
fclose($fp);
        } else {
            
$this->cakeError('randomNumber');
        }
        
        
$time_low bin2hex(substr($pr_bits,04));
        
$time_mid bin2hex(substr($pr_bits,42));
        
$time_hi_and_version bin2hex(substr($pr_bits,62));
        
$clock_seq_hi_and_reserved bin2hex(substr($pr_bits,82));
        
$node bin2hex(substr($pr_bits,106));
        
        

        
$time_hi_and_version hexdec($time_hi_and_version);
        
$time_hi_and_version $time_hi_and_version >> 4;
        
$time_hi_and_version $time_hi_and_version 0x4000;
        
        

        
$clock_seq_hi_and_reserved hexdec($clock_seq_hi_and_reserved);
        
$clock_seq_hi_and_reserved $clock_seq_hi_and_reserved >> 2;
        
$clock_seq_hi_and_reserved $clock_seq_hi_and_reserved 0x8000;
        
        return 
sprintf('%08s-%04s-%04x-%04x-%012s',
            
$time_low$time_mid$time_hi_and_version$clock_seq_hi_and_reserved$node);
    }
?>

[#20] mark at whytewaters dot com [2008-02-25 18:39:10]

If you want many ids and performance of this function is an issue why not pull uniquid() out of the loop, eg:

$base = uniqueid();
$ids[] = array();

for ($index = 0; $index < 100000; $index++)  
    $ids[] = $base . '.' . $index;

[#21] nodkz at mail dot ru [2008-01-15 02:00:57]

I use such UUID (it not RFC!!!)
(server_id)-(clientIP)-(unixtime)-(milliseconds)-(random)

 I can easyly determine which server at which time and who initiate creating of object.

<?php

$u
=uuid();   // 0001-7f000001-478c8000-4801-47242987
echo $u;
echo 
"<br>";
print_r(uuidDecode($u)); // Array ( [serverID] => 0001 [ip] => 127.0.0.1 [unixtime] => 1200390144 [micro] => 0.28126525878906 ) 

function uuid($serverID=1)
{
    
$t=explode(" ",microtime());
    return 
sprintf'%04x-%08s-%08s-%04s-%04x%04x',
        
$serverID,
        
clientIPToHex(),
        
substr("00000000".dechex($t[1]),-8),   // get 8HEX of unixtime
        
substr("0000".dechex(round($t[0]*65536)),-4), // get 4HEX of microtime
        
mt_rand(0,0xffff), mt_rand(0,0xffff));
}

function 
uuidDecode($uuid) {
    
$rez=Array();
    
$u=explode("-",$uuid);
    if(
is_array($u)&&count($u)==5) {
        
$rez=Array(
            
'serverID'=>$u[0],
            
'ip'=>clientIPFromHex($u[1]),
            
'unixtime'=>hexdec($u[2]),
            
'micro'=>(hexdec($u[3])/65536)
        );
    }
    return 
$rez;
}

function 
clientIPToHex($ip="") {
    
$hex="";
    if(
$ip==""$ip=getEnv("REMOTE_ADDR");
    
$part=explode('.'$ip);
    for (
$i=0$i<=count($part)-1$i++) {
        
$hex.=substr("0".dechex($part[$i]),-2);
    }
    return 
$hex;
}

function 
clientIPFromHex($hex) {
    
$ip="";
    if(
strlen($hex)==8) {
        
$ip.=hexdec(substr($hex,0,2)).".";
        
$ip.=hexdec(substr($hex,2,2)).".";
        
$ip.=hexdec(substr($hex,4,2)).".";
        
$ip.=hexdec(substr($hex,6,2));
    }
    return 
$ip;
}

?>

[#22] ken at smallboxsoftware [2007-05-17 09:34:16]

Just to note this function is fairly slow, and can bring your script to a crawl if it is in a loop. Strangely if you run it as uniqid('', true) it runs much more quickly

[#23] Jason [2007-03-27 12:10:36]

Neither the pseudo-random number rand() nor the Mersenne Twister algorithms are cryptographically strong, and this is well known.  Simply combining non-cryptographically strong algorithms doesn't not make a cryptographically strong algorithm either.  Mersenne Twister is a fast algorithm with good k-distribution which will give you numbers for a long time before it repeats itself.  MT, rand(), and MD5 should NOT be used for encryption, or for cookies that that store a session ID that gives personal information.  A simple application where non-collision of session IDs is highly preferred but not critical, such as storing a user's shopping cart items for when they return to your site (but not their personal information), IS a good use for the MT, rand() MD5, uniqid() and combinations thereof.

[#24] mailrinke at _cutthis_yahoo dot com [2007-02-19 05:06:44]

I have been using mimecs version lately and do not think it's safe to think the results are always unqiue.

Although it could be just my bad programming, I found exactly 1 collission while debugging my code. It seems to me that if my code was incorrect it would have happened more than once.

I recommend anyone to include time as a factor of such an ID as to be a little more certain it is in fact unique.

[#25] Emery [2007-01-31 00:13:40]

The example given in this document for a "better token" should be:

<?php
$better_token 
uniqid(md5(rand()), true);
?>


As it is now, the result isn't guaranteed to be unique, because MD5 has collisions.

[#26] lance_rushing at hotmail dot com [2007-01-23 12:43:49]

wooshoofoo, the reason mimec is calling mt_rand multiple times is because the largest number mt_rand can produce is 2^31 (2147483647, as reported by mt_getrandmax() on my server).  RFC 4122 requires a 128 bit value.

Also they are not "4 digit sequeces", but 4 digit hexadecimal numbers.  16^4 == 2^16.

mimec's limiting each random result to 2^16 avoids problem of PHP's 2^32 integer max (http://php.net/manual/en/language.types.integer.php).

If you want to call mt_rand fewer times:  mimec's version calls mt_rand 8 times ( 16 bits * 8 = 128 bits ).  You *could* call mt_rand 5 times ( 31 bits + 31 bits + 31 bits + 31 bits + 4 bits = 128 bits ).  But then you would have keep all your values as strings.

Something like: 

<?php

function uuid() {    

     
// Generate 128 bit random sequence 
     
$randmax_bits strlen(base_convert(mt_getrandmax(), 102));  // how many bits is mt_getrandmax()
     
$x '';
     while (
strlen($x) < 128) {
         
$maxbits = (128 strlen($x) < $randmax_bits) ? 128 strlen($x) :  $randmax_bits;
         
$x .= str_pad(base_convert(mt_rand(0pow(2,$maxbits)), 102), $maxbits"0"STR_PAD_LEFT);
     }

     
// break into fields
     
$a = array();
     
$a['time_low_part'] = substr($x032);
     
$a['time_mid'] = substr($x3216);
     
$a['time_hi_and_version'] = substr($x4816);
     
$a['clock_seq'] = substr($x6416);
     
$a['node_part'] =  substr($x8048);
     
     
// Apply bit masks for "random or pseudo-random" version per RFC
     
$a['time_hi_and_version'] = substr_replace($a['time_hi_and_version'], '0100'04);
     
$a['clock_seq'] = substr_replace($a['clock_seq'], '10'02);

    
// Format output
    
return sprintf('%s-%s-%s-%s-%s',
        
str_pad(base_convert($a['time_low_part'], 216), 8"0"STR_PAD_LEFT),
        
str_pad(base_convert($a['time_mid'], 216), 4"0"STR_PAD_LEFT),
        
str_pad(base_convert($a['time_hi_and_version'], 216), 4"0"STR_PAD_LEFT),
        
str_pad(base_convert($a['clock_seq'], 216), 4"0"STR_PAD_LEFT),
        
str_pad(base_convert($a['node_part'], 216), 12"0"STR_PAD_LEFT));
}

?>


However, I think mimec's version is much more elegant.

[#27] wooshoofoo [2006-12-06 18:00:04]

I'm not sure the previous function by mimec is really all that random.  For one thing, generating 8 small random 4 digit sequeces != generating one 32 digit sequence.

[#28] admin at code-dynasty dot net [2006-07-08 20:46:19]

I'm not too fond of the recommendation to use an MD5 of the unique ID for session IDs. It would be a better idea just to use uniqueid(rand(), true) without the MD5, because even though it's a rare circumstance, MD5 is a hash, not an encryption, which means it has collisions. Therefore you theoretically could have multiple users given the same session ID which could result in one user's ability to access another user's data.

上一篇: 下一篇: