Keep in mind that the Suhosin patch which is installed by default on many PHP-installs such as Debian and DirectAdmin completely disables the srand and mt_srand functions for encryption security reasons. To generate reproducible random numbers from a fixed seed on a Suhosin-hardened server you will need to include your own pseudorandom generator code.
srand
(PHP 4, PHP 5, PHP 7, PHP 8)
srand — Seed the random number generator
Descripción
Seeds the random number generator with seed
or with a random value if seed is 0.
Nota: No es necesario usar una semilla para usar el generador de números aleatorios con srand() o mt_srand() ya que se hace automáticamente.
Precaución
Because the Mt19937 (“Mersenne Twister”) engine accepts only a single 32 bit integer as the seed, the number of possible random sequences is limited to just 232 (i.e. 4,294,967,296), despite Mt19937’s huge period of 219937-1.
When relying on either implicit or explicit random seeding, duplications will appear much earlier. Duplicated seeds are expected with 50% probability after less than 80,000 randomly generated seeds according to the birthday problem. A 10% probability of a duplicated seed happens after randomly generating roughly 30,000 seeds.
This makes Mt19937 unsuitable for applications where duplicated sequences must not happen with more than a negligible probability. If reproducible seeding is required, both the Random\Engine\Xoshiro256StarStar and Random\Engine\PcgOneseq128XslRr64 engines support much larger seeds that are unlikely to collide randomly. If reproducibility is not required, the Random\Engine\Secure engine provides cryptographically secure randomness.
Nota: As of PHP 7.1.0, srand() has been made an alias of mt_srand().
Parámetros
-
seed -
Fills the state with values generated with a linear congruential generator that was seeded with
seedinterpreted as an unsigned 32 bit integer.If
seedis omitted ornull, a random unsigned 32-bit integer will be used.
Valores devueltos
No devuelve ningún valor.
Historial de cambios
| Versión | Descripción |
|---|---|
| 8.3.0 |
seed is now nullable.
|
| 7.1.0 | srand() has been made an alias of mt_srand(). |
Ver también
- rand() - Generate a random integer
- getrandmax() - Show largest possible random value
- mt_srand() - Seeds the Mersenne Twister Random Number Generator
add a note
User Contributed Notes 13 notes
Niels Keurentjes ¶
13 years ago
harmen at no dot spam dot rdzl dot nl ¶
15 years ago
To generate a random number which is different every day, I used the number of days after unix epoch as a seed:
<?php
srand(floor(time() / (60*60*24)));
echo rand() % 100;
?>
My provider upgraded the php server recently, and calling srand(seed) does not seem to set the seed anymore. To let srand set the seed, add the following line to your .htaccess file
php_value suhosin.srand.ignore 0
Kudos to doc_z (http://www.webmasterworld.com/php/3777515.htm)
Harmen
rjones at ditzed dot org ¶
23 years ago
Use the srand() seed "(double)microtime()*1000000" as mentioned by the richard@zend.com at the top of these user notes.
The most notable effect of using any other seed is that your random numbers tend to follow the same, or very similar, sequences each time the script is invoked.
Take note of the following script:
<?php
srand($val);
echo rand(0, 20) . ", ";
echo rand(0, 20) . ", ";
echo rand(0, 20) . ", ";
echo rand(0, 20) . ", ";
echo rand(0, 20);
?>
If you seed the generator with a constant, say; the number 5 ($val = 5), then the sequence generated is always the same, in this case (0, 18, 7, 15, 17) (for me at least, different processors/processor speeds/operating systems/OS releases/PHP releases/webserver software may generate different sequences).
If you seed the generator with time(), then the sequence is more random, but invokations that are very close together will have similar outputs.
As richard@zend.com above suggests, the best seed to use is (double) microtime() * 1000000, as this gives the greatest amount of psuedo-randomness. In fact, it is random enough to suit most users.
In a test program of 100000 random numbers between 1 and 20, the results were fairly balanced, giving an average of 5000 results per number, give or take 100. The deviation in each case varied with each invokation.
edublancoa at gmail dot com ¶
19 years ago
Another use of srand is to obtain the same value of rand in a determined time interval. Example: you have an array of 100 elements and you need to obtain a random item every day but not to change in the 24h period (just imagine "Today's Photo" or similar).
<?php
$seed = floor(time()/86400);
srand($seed);
$item = $examplearray[rand(0,99)];
?>
You obtain the same value every time you load the page all the 24h period.
contact at einenlum dot com ¶
10 months ago
Keep in mind that now srand is an alias for mt_srand, but they behaved differently before. This means you should not follow the documentation of srand, but the one of mt_srand, when using srand.
To reset the seed to a random value, `mt_srand(0)` (or `srand(0)`) doesn't work. It sets the seed to 0. To reset the seed to a random value you must use `mt_srand()` (or `srand()`).
<?php
$arr = [0, 1, 2, 3, 4];
srand(1); // or mt_srand(1) as they are now aliases
$keys = array_rand($arr, 2); // not random as expected
srand(0); // or mt_srand(0) as they are now aliases
$keys = array_rand($arr, 2); // not random either!
srand(); // or mt_srand() as they are now aliases
$keys = array_rand($arr, 2); // random again
?>
Glauco Lins ¶
9 years ago
srand and mt_srand are both initialized only once per process ID.
You cannot re-seed your rand algorithms after the first "srand", "mt_srand", "rand", "mt_rand", "shuffle", or any other rand-like function.
I have been facing an issue where after forking my process, all childs were generating exactly the same rand values.
This was due a first "shuffle" call on the parent process, so I could not re-seed the childs.
To solve my issue, I simple called "rand" N times, to offset the child rand generators.
# Offset the child rand generator by its PID
$n = (getmypid() % 100) * (10 * abs(microtime(true) - time()));
for ($n; $n > 0; $n--) {
rand(0, $n);
}
Since each pcntl_fork takes a while to be completed, the microtime offers an extra offset, other than one PID increment.
This small code will make at the WORST hypothesis 1000 iterations.
mlwmohawk at mohawksoft dot com ¶
23 years ago
srand() is pretty tricky to get right. You should never seed a random number generator more than once per php process, if you do, your randomness is limited to the source of your seed.
The microtime function's micro-seconds portion has a very finite resolution, that is why the make_seed function was added to the document. You should never get the same seed twice.
In the later CVS versions, PHP will seed the random generator prior to performing a rand() if srand() was not previously called.
Anonymous ¶
23 years ago
I have a ramdon circulater that changes a piece of text once a day, and I use the following code to make sure the see is unique enough.
$tm = time();
$today = mktime(0, 0, 0, (int)date("n", $tm), (int)date("j", $tm), (int)date("Y", $tm));
srand($today / pi());
The pi works wonders for the whole thing and it works like a charm. Any other big decimal number will do as well, but pi is the most common "big" number.
rjones at ditzed dot org ¶
23 years ago
As a sidenote on the usage of srand():
If you are making use of modular programming, it is probably better to try and call the srand routine from the parent script than from any modules you may be using (using REQUIRE or INCLUDE).
This way you get around the possibility of calling srand() more than once from different modules.
The flaw in this solution, of course, is when using modules produced by another programmer, or when producing modules for another programmer.
You cannot rely on another programmer calling the srand function before calling the modular function, so you would have to include the srand function inside the module in this case.
If you produce modules for use by other programmers then it is good practice to documentise the fact you have already called the srand function.
Or if you use a modular function produced by someone else, check their documentation, or check their source code.
MakeMoolah at themail dot com ¶
23 years ago
Sorry about that... ok, forget have of what I said up there ^.
The code that would prove my example is this:
<?php
srand(5);
echo(rand(1, 10));
srand(5);
echo(rand(1, 10));
srand(5);
echo(rand(1, 10));
?>
Each time you SHOULD get the same answer, but if you did this:
<?php
srand(5);
echo(rand(1, 10));
echo(rand(1, 10));
echo(rand(1, 10));
?>
then the answers would be different, and you'd be letting the random number formula do it's duty.
hagen at von-eitzen dot de ¶
24 years ago
It is REALLY essential to make sure that srand is called only once.
This is a bit difficult if the call is hidden somewhere in third-party code you include. For example, I used a standard banner script that *seemed* to work well putting
three random banners in one frame. But in the long run, the choice appeared
somewhat biased - probably because srand was called once per banner, not
once per run.
It would be nice if the random number generator worked like in PERL: If You use the random function without having called srand ever before in a script,
srand is invoked before (and automatically with a nice seed, hopefully).
I suggest that one should do something like this:
<?php
if (!$GLOBALS["IHaveCalledSrandBefore"]++) {
srand((double) microtime() * 1000000);
}
?>
(Depending on the situation, one might also work with a static variable instead)
bootc at bootc dot net ¶
19 years ago
OK, to summarize what people have been saying so far:
1. DO NOT seed the RNG more than once if you can help it!
2. You HAVE TO seed the RNG yourself if you are using PHP < 4.2.0.
3. Using a prime multiplier to microtime() probably does very little. Use the Mersenne Twister instead.
4. You can use the Mersenne Twister PRNG with the mt_rand and mt_srand functions. This is faster and is more random.
akukula at min dot pl ¶
23 years ago
Calling srand((double)microtime()*1000000),
then $a=rand(1000000,9999999), then srand((double)microtime()*$a)
adds nothing to the entrophy: the execution time of rand and srand is
constant, so the second microtime() produces nothing really fascinating. You may safely use just the first srand().
