Related
How does one generate a random float between 0 and 1 in PHP?
I'm looking for the PHP's equivalent to Java's Math.random().
You may use the standard function: lcg_value().
Here's another function given on the rand() docs:
// auxiliary function
// returns random number with flat distribution from 0 to 1
function random_0_1()
{
return (float)rand() / (float)getrandmax();
}
Example from documentation :
function random_float ($min,$max) {
return ($min+lcg_value()*(abs($max-$min)));
}
rand(0,1000)/1000 returns:
0.348 0.716 0.251 0.459 0.893 0.867 0.058 0.955 0.644 0.246 0.292
or use a bigger number if you want more digits after decimal point
class SomeHelper
{
/**
* Generate random float number.
*
* #param float|int $min
* #param float|int $max
* #return float
*/
public static function rand($min = 0, $max = 1)
{
return ($min + ($max - $min) * (mt_rand() / mt_getrandmax()));
}
}
update:
forget this answer it doesnt work wit php -v > 5.3
What about
floatVal('0.'.rand(1, 9));
?
this works perfect for me, and it´s not only for 0 - 1 for example between 1.0 - 15.0
floatVal(rand(1, 15).'.'.rand(1, 9));
function mt_rand_float($min, $max, $countZero = '0') {
$countZero = +('1'.$countZero);
$min = floor($min*$countZero);
$max = floor($max*$countZero);
$rand = mt_rand($min, $max) / $countZero;
return $rand;
}
example:
echo mt_rand_float(0, 1);
result: 0.2
echo mt_rand_float(3.2, 3.23, '000');
result: 3.219
echo mt_rand_float(1, 5, '00');
result: 4.52
echo mt_rand_float(0.56789, 1, '00');
result: 0.69
$random_number = rand(1,10).".".rand(1,9);
function frand($min, $max, $decimals = 0) {
$scale = pow(10, $decimals);
return mt_rand($min * $scale, $max * $scale) / $scale;
}
echo "frand(0, 10, 2) = " . frand(0, 10, 2) . "\n";
This question asks for a value from 0 to 1. For most mathematical purposes this is usually invalid albeit to the smallest possible degree. The standard distribution by convention is 0 >= N < 1. You should consider if you really want something inclusive of 1.
Many things that do this absent minded have a one in a couple billion result of an anomalous result. This becomes obvious if you think about performing the operation backwards.
(int)(random_float() * 10) would return a value from 0 to 9 with an equal chance of each value. If in one in a billion times it can return 1 then very rarely it will return 10 instead.
Some people would fix this after the fact (to decide that 10 should be 9). Multiplying it by 2 should give around a ~50% chance of 0 or 1 but will also have a ~0.000000000465% chance of returning a 2 like in Bender's dream.
Saying 0 to 1 as a float might be a bit like mistakenly saying 0 to 10 instead of 0 to 9 as ints when you want ten values starting at zero. In this case because of the broad range of possible float values then it's more like accidentally saying 0 to 1000000000 instead of 0 to 999999999.
With 64bit it's exceedingly rare to overflow but in this case some random functions are 32bit internally so it's not no implausible for that one in two and a half billion chance to occur.
The standard solutions would instead want to be like this:
mt_rand() / (getrandmax() + 1)
There can also be small usually insignificant differences in distribution, for example between 0 to 9 then you might find 0 is slightly more likely than 9 due to precision but this will typically be in the billionth or so and is not as severe as the above issue because the above issue can produce an invalid unexpected out of bounds figure for a calculation that would otherwise be flawless.
Java's Math.random will also never produce a value of 1. Some of this comes from that it is a mouthful to explain specifically what it does. It returns a value from 0 to less than one. It's Zeno's arrow, it never reaches 1. This isn't something someone would conventionally say. Instead people tend to say between 0 and 1 or from 0 to 1 but those are false.
This is somewhat a source of amusement in bug reports. For example, any PHP code using lcg_value without consideration for this may glitch approximately one in a couple billion times if it holds true to its documentation but that makes it painfully difficult to faithfully reproduce.
This kind of off by one error is one of the common sources of "Just turn it off and on again." issues typically encountered in embedded devices.
Solution for PHP 7. Generates random number in [0,1). i.e. includes 0 and excludes 1.
function random_float() {
return random_int(0, 2**53-1) / (2**53);
}
Thanks to Nommyde in the comments for pointing out my bug.
>>> number_format((2**53-1)/2**53,100)
=> "0.9999999999999998889776975374843459576368331909179687500000000000000000000000000000000000000000000000"
>>> number_format((2**53)/(2**53+1),100)
=> "1.0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000"
Most answers are using mt_rand. However, mt_getrandmax() usually returns only 2147483647. That means you only have 31 bits of information, while a double has a mantissa with 52 bits, which means there is a density of at least 2^53 for the numbers between 0 and 1.
This more complicated approach will get you a finer distribution:
function rand_754_01() {
// Generate 64 random bits (8 bytes)
$entropy = openssl_random_pseudo_bytes(8);
// Create a string of 12 '0' bits and 52 '1' bits.
$x = 0x000FFFFFFFFFFFFF;
$first12 = pack("Q", $x);
// Set the first 12 bits to 0 in the random string.
$y = $entropy & $first12;
// Now set the first 12 bits to be 0[exponent], where exponent is randomly chosen between 1 and 1022.
// Here $e has a probability of 0.5 to be 1022, 0.25 to be 1021, etc.
$e = 1022;
while($e > 1) {
if(mt_rand(0,1) == 0) {
break;
} else {
--$e;
}
}
// Pack the exponent properly (add four '0' bits behind it and 49 more in front)
$z = "\0\0\0\0\0\0" . pack("S", $e << 4);
// Now convert to a double.
return unpack("d", $y | $z)[1];
}
Please note that the above code only works on 64-bit machines with a Litte-Endian byte order and Intel-style IEEE754 representation. (x64-compatible computers will have this). Unfortunately PHP does not allow bit-shifting past int32-sized boundaries, so you have to write a separate function for Big-Endian.
You should replace this line:
$z = "\0\0\0\0\0\0" . pack("S", $e << 4);
with its big-endian counterpart:
$z = pack("S", $e << 4) . "\0\0\0\0\0\0";
The difference is only notable when the function is called a large amount of times: 10^9 or more.
Testing if this works
It should be obvious that the mantissa follows a nice uniform distribution approximation, but it's less obvious that a sum of a large amount of such distributions (each with cumulatively halved chance and amplitude) is uniform.
Running:
function randomNumbers() {
$f = 0.0;
for($i = 0; $i < 1000000; ++$i) {
$f += \math::rand_754_01();
}
echo $f / 1000000;
}
Produces an output of 0.49999928273099 (or a similar number close to 0.5).
I found the answer on PHP.net
<?php
function randomFloat($min = 0, $max = 1) {
return $min + mt_rand() / mt_getrandmax() * ($max - $min);
}
var_dump(randomFloat());
var_dump(randomFloat(2, 20));
?>
float(0.91601131712832)
float(16.511210331931)
So you could do
randomFloat(0,1);
or simple
mt_rand() / mt_getrandmax() * 1;
what about:
echo (float)('0.' . rand(0,99999));
would probably work fine... hope it helps you.
I'm trying to write an algorithm to convert a 64Bit integer to 32Bit via truncation, and return a value accurately representing a 32Bit value. The obvious problem is that integers in PHP are only 64Bit (barring 32Bit systems).
I first tried the following:
$x = $v & 0xFFFFFFFF
So if $v = PHP_INT_MAX, $x is 4294967295, when really I want -1. I know the reason for this is that when represented as a 64Bit integer, PHP prepends the zeros to the last 32 bits, and thats why I get a positive number.
My solution so far is this:
function convert64BitTo32Bit(int $v): int
{
$v &= 0xFFFFFFFF;
if ($v & 0x80000000) {
return $v | 0xFFFFFFFF00000000;
}
return $v;
}
And I'm pretty sure its right. The problem I have with this is that it requires that if statement to inspect whether the truncated number is negative. I was really hoping for a bitwise only solution. It may not be possible, but I thought I'd ask.
EDIT:
The problem can be simplified to just part of the solution. i.e. if the first bit is 1, make all bits 1, if first bit is 0 make all bits 0.
# example assumes input is the LSBs of an 8Bit integer.
# scale for 64Bit and the same solution should work.
op(1010) = 1111
op(0101) = 0000
op(0000) = 0000
op(1000) = 1111
op(0001) = 0000
I would be able to use the LSBs to derive this value, then mask it onto the MSBs of the 64Bit integer. This is what I'm trying to figure out now, though I'm trying to avoid creating a monster equation.
There's probably a more elegant/efficient way using bitwise operations, but you can also force the conversion with pack() and unpack(). Eg: Pack as unsigned, unpack as signed.
function overflow32($in) {
return unpack('l', pack('i', $in & 0xFFFFFFFF))[1];
}
var_dump( overflow32(pow(2,33)-1) ); // int(-1)
I'm curious what you're applying this to, because I had to break a hash function with this same thing some years ago when I moved an app from a 32 bit machine to a 64 bit machine, but I can't remember what it was.
Edit: Wow for some reason I remembered Two's Complement being hard. Literally just invert and add one.
function overflow32($in) {
if( $in & 0x80000000 ) {
return ( ( ~$in & 0xFFFFFFFF) + 1 ) * -1;
} else {
return $in & 0xFFFFFFFF;
}
}
var_dump( kludge32(pow(2,33)-1) );
Edit 2: I saw that you want to extend this to arbitrary bit lengths, in which case you just need to calculate the masks instead of explicitly setting them:
function overflow_bits($in, $bits=32) {
$sign_mask = 1 << $bits-1;
$clamp_mask = ($sign_mask << 1) - 1;
var_dump(
decbin($in),
decbin($sign_mask),
decbin($clamp_mask)
);
if( $in & $sign_mask ) {
return ( ( ~$in & $clamp_mask) + 1 ) * -1;
} else {
return $in & $clamp_mask;
}
}
var_dump(
overflow_bits(pow(2, 31), 32),
overflow_bits(pow(2, 15), 16),
overflow_bits(pow(2, 7), 8)
);
I left in the debug var_dump()s for output flavor:
string(32) "10000000000000000000000000000000"
string(32) "10000000000000000000000000000000"
string(32) "11111111111111111111111111111111"
string(16) "1000000000000000"
string(16) "1000000000000000"
string(16) "1111111111111111"
string(8) "10000000"
string(8) "10000000"
string(8) "11111111"
int(-2147483648)
int(-32768)
int(-128)
I think I've cracked it:
function overflow32Bit(int $x): int
{
return ($x & 0xFFFFFFFF) | ((($x & 0xFFFFFFFF) >> 31) * ((2 ** 32) - 1) << 32);
}
var_dump(overflow32Bit(PHP_INT_MAX)); // -1 correct
var_dump(overflow32Bit(PHP_INT_MAX - 1)); // -2 correct
var_dump(overflow32Bit(PHP_INT_MIN)); // 0 correct
var_dump(overflow32Bit((2 ** 31) - 1)); // 2147483647 correct
var_dump(overflow32Bit((2 ** 31))); // -2147483647 correct
var_dump(overflow32Bit(0xFFFFFFFF)); // -1 correct
var_dump(overflow32Bit(0x7FFFFFFF)); // 2147483647 correct
The solution was actually staring me in the face. Get the value of the first bit, then multiply it by the max value of unsigned 32bit integer.
If someone can come up with a better or shorter solution, I'll also accept that.
PS. This is only for 32Bit, i also intend to use this proof for 16Bit and 8Bit.
Expanded, $x is input, $z is output.
$x = PHP_INT_MAX;
$y = (2 ** 32) - 1;
$z = ($x & $y) | ((($x & $y) >> 31) * ($y << 32));
var_dump($z);
I'm trying to convert a 64-bit float to a 64-bit integer (and back) in php. I need to preserve the bytes, so I'm using the pack and unpack functions. The functionality I'm looking for is basically Java's Double.doubleToLongBits() method. http://docs.oracle.com/javase/7/docs/api/java/lang/Double.html#doubleToLongBits(double)
I managed to get this far with some help from the comments on the php docs for pack():
function encode($int) {
$int = round($int);
$left = 0xffffffff00000000;
$right = 0x00000000ffffffff;
$l = ($int & $left) >>32;
$r = $int & $right;
return unpack('d', pack('NN', $l, $r))[1];
}
function decode($float) {
$set = unpack('N2', pack('d', $float));
return $set[1] << 32 | $set[2];
}
And this works well, for the most part...
echo decode(encode(10000000000000));
100000000
echo encode(10000000000000);
1.1710299640683E-305
But here's where it gets tricky...
echo decode(1.1710299640683E-305);
-6629571225977708544
I have no idea what's wrong here. Try it for yourself: http://pastebin.com/zWKC97Z7
You'll need 64-bit PHP on linux. This site seems to emulate that setup: http://www.compileonline.com/execute_php_online.php
$x = encode(10000000000000);
var_dump($x); //float(1.1710299640683E-305)
echo decode($x); //10000000000000
$y = (float) "1.1710299640683E-305";
var_dump($y); //float(1.1710299640683E-305)
echo decode($y); //-6629571225977708544
$z = ($x == $y);
var_dump($z); //false
http://www.php.net/manual/en/language.types.float.php
... never trust
floating number results to the last digit, and do not compare floating
point numbers directly for equality. If higher precision is necessary,
the arbitrary precision math functions and gmp functions are
available. For a "simple" explanation, see the » floating point guide
that's also titled "Why don’t my numbers add up?"
It is working properly, the only problem in this case is in logic of:
echo decode(1.1710299640683E-305);
You can't use "rounded" and "human readable" output of echo function to decode the original value (because you are loosing precision of this double then).
If you will save the return of encode(10000000000000) to the variable and then try to decode it again it will works properly (you can use echo on 10000000000000 without loosing precision).
Please see the example below which you can execute on PHP compiler as well:
<?php
function encode($int) {
$int = round($int);
$left = 0xffffffff00000000;
$right = 0x00000000ffffffff;
$l = ($int & $left) >>32;
$r = $int & $right;
return unpack('d', pack('NN', $l, $r))[1];
}
function decode($float) {
$set = unpack('N2', pack('d', $float));
return $set[1] << 32 | $set[2];
}
echo decode(encode(10000000000000)); // untouched
echo '<br /><br />';
$encoded = encode(10000000000000);
echo $encoded; // LOOSING PRECISION!
echo ' - "human readable" version of encoded int<br /><br />';
echo decode($encoded); // STILL WORKS - HAPPY DAYS!
?>
If you have a reliable fixed decimal point, like in my case and the case of currency, you can multiply your float by some power of 10 (ex. 100 for dollars).
function encode($float) {
return (int) $float * pow(10, 2);
}
function decode($str) {
return bcdiv($str, pow(10, 2), 2);
}
However, this doesn't work for huge numbers and doesn't officially solve the problem.
Seems like it's impossible to convert from an integer to a float string and back without losing the original integer value in php 5.4
I wonder if is there a good way to get the number of digits in right/left side of a decimal number PHP. For example:
12345.789 -> RIGHT SIDE LENGTH IS 3 / LEFT SIDE LENGTH IS 5
I know it is readily attainable by helping string functions and exploding the number. I mean is there a mathematically or programmatically way to perform it better than string manipulations.
Your answers would be greatly appreciated.
Update
The best solution for left side till now was:
$left = floor(log10($x))+1;
but still no sufficient for right side.
Still waiting ...
To get the digits on the left side you can do this:
$left = floor(log10($x))+1;
This uses the base 10 logarithm to get the number of digits.
The right side is harder. A simple approach would look like this, but due to floating point numbers, it would often fail:
$decimal = $x - floor($x);
$right = 0;
while (floor($decimal) != $decimal) {
$right++;
$decimal *= 10; //will bring in floating point 'noise' over time
}
This will loop through multiplying by 10 until there are no digits past the decimal. That is tested with floor($decimal) != $decimal.
However, as Ali points out, giving it the number 155.11 (a hard to represent digit in binary) results in a answer of 14. This is because as the number is stored as something like 155.11000000000001 with the 32 bits of floating precision we have.
So instead, a more robust solution is needed. (PoPoFibo's solutions above is particularly elegant, and uses PHPs inherit float comparison functions well).
The fact is, we can never distinguish between input of 155.11 and 155.11000000000001. We will never know which number was originally given. They will both be represented the same. However, if we define the number of zeroes that we can see in a row before we just decide the decimal is 'done' than we can come up with a solution:
$x = 155.11; //the number we are testing
$LIMIT = 10; //number of zeroes in a row until we say 'enough'
$right = 0; //number of digits we've checked
$empty = 0; //number of zeroes we've seen in a row
while (floor($x) != $x) {
$right++;
$base = floor($x); //so we can see what the next digit is;
$x *= 10;
$base *= 10;
$digit = floor($x) - $base; //the digit we are dealing with
if ($digit == 0) {
$empty += 1;
if ($empty == $LIMIT) {
$right -= $empty; //don't count all those zeroes
break; // exit the loop, we're done
}
} else {
$zeros = 0;
}
}
This should find the solution given the reasonable assumption that 10 zeroes in a row means any other digits just don't matter.
However, I still like PopoFibo's solution better, as without any multiplication, PHPs default comparison functions effectively do the same thing, without the messiness.
I am lost on PHP semantics big time but I guess the following would serve your purpose without the String usage (that is at least how I would do in Java but hopefully cleaner):
Working code here: http://ideone.com/7BnsR3
Non-string solution (only Math)
Left side is resolved hence taking the cue from your question update:
$value = 12343525.34541;
$left = floor(log10($value))+1;
echo($left);
$num = floatval($value);
$right = 0;
while($num != round($num, $right)) {
$right++;
}
echo($right);
Prints
85
8 for the LHS and 5 for the RHS.
Since I'm taking a floatval that would make 155.0 as 0 RHS which I think is valid and can be resolved by String functions.
php > $num = 12345.789;
php > $left = strlen(floor($num));
php > $right = strlen($num - floor($num));
php > echo "$left / $right\n";
5 / 16 <--- 16 digits, huh?
php > $parts = explode('.', $num);
php > var_dump($parts);
array(2) {
[0]=>
string(5) "12345"
[1]=>
string(3) "789"
As you can see, floats aren't the easiest to deal with... Doing it "mathematically" leads to bad results. Doing it by strings works, but makes you feel dirty.
$number = 12345.789;
list($whole, $fraction) = sscanf($number, "%d.%d");
This will always work, even if $number is an integer and you’ll get two real integers returned. Length is best done with strlen() even for integer values. The proposed log10() approach won't work for 10, 100, 1000, … as you might expect.
// 5 - 3
echo strlen($whole) , " - " , strlen($fraction);
If you really, really want to get the length without calling any string function here you go. But it's totally not efficient at all compared to strlen().
/**
* Get integer length.
*
* #param integer $integer
* The integer to count.
* #param boolean $count_zero [optional]
* Whether 0 is to be counted or not, defaults to FALSE.
* #return integer
* The integer's length.
*/
function get_int_length($integer, $count_zero = false) {
// 0 would be 1 in string mode! Highly depends on use case.
if ($count_zero === false && $integer === 0) {
return 0;
}
return floor(log10(abs($integer))) + 1;
}
// 5 - 3
echo get_int_length($whole) , " - " , get_int_length($fraction);
The above will correctly count the result of 1 / 3, but be aware that the precision is important.
$number = 1 / 3;
// Above code outputs
// string : 1 - 10
// math : 0 - 10
$number = bcdiv(1, 3);
// Above code outputs
// string : 1 - 0 <-- oops
// math : 0 - INF <-- 8-)
No problem there.
I would like to apply a simple logic.
<?php
$num=12345.789;
$num_str="".$num; // Converting number to string
$array=explode('.',$num_str); //Explode number (String) with .
echo "Left side length : ".intval(strlen($array[0])); // $array[0] contains left hand side then check the string length
echo "<br>";
if(sizeof($array)>1)
{
echo "Left side length : ".intval(strlen($array[1]));// $array[1] contains left hand check the string length side
}
?>
How does one generate a random float between 0 and 1 in PHP?
I'm looking for the PHP's equivalent to Java's Math.random().
You may use the standard function: lcg_value().
Here's another function given on the rand() docs:
// auxiliary function
// returns random number with flat distribution from 0 to 1
function random_0_1()
{
return (float)rand() / (float)getrandmax();
}
Example from documentation :
function random_float ($min,$max) {
return ($min+lcg_value()*(abs($max-$min)));
}
rand(0,1000)/1000 returns:
0.348 0.716 0.251 0.459 0.893 0.867 0.058 0.955 0.644 0.246 0.292
or use a bigger number if you want more digits after decimal point
class SomeHelper
{
/**
* Generate random float number.
*
* #param float|int $min
* #param float|int $max
* #return float
*/
public static function rand($min = 0, $max = 1)
{
return ($min + ($max - $min) * (mt_rand() / mt_getrandmax()));
}
}
update:
forget this answer it doesnt work wit php -v > 5.3
What about
floatVal('0.'.rand(1, 9));
?
this works perfect for me, and it´s not only for 0 - 1 for example between 1.0 - 15.0
floatVal(rand(1, 15).'.'.rand(1, 9));
function mt_rand_float($min, $max, $countZero = '0') {
$countZero = +('1'.$countZero);
$min = floor($min*$countZero);
$max = floor($max*$countZero);
$rand = mt_rand($min, $max) / $countZero;
return $rand;
}
example:
echo mt_rand_float(0, 1);
result: 0.2
echo mt_rand_float(3.2, 3.23, '000');
result: 3.219
echo mt_rand_float(1, 5, '00');
result: 4.52
echo mt_rand_float(0.56789, 1, '00');
result: 0.69
$random_number = rand(1,10).".".rand(1,9);
function frand($min, $max, $decimals = 0) {
$scale = pow(10, $decimals);
return mt_rand($min * $scale, $max * $scale) / $scale;
}
echo "frand(0, 10, 2) = " . frand(0, 10, 2) . "\n";
This question asks for a value from 0 to 1. For most mathematical purposes this is usually invalid albeit to the smallest possible degree. The standard distribution by convention is 0 >= N < 1. You should consider if you really want something inclusive of 1.
Many things that do this absent minded have a one in a couple billion result of an anomalous result. This becomes obvious if you think about performing the operation backwards.
(int)(random_float() * 10) would return a value from 0 to 9 with an equal chance of each value. If in one in a billion times it can return 1 then very rarely it will return 10 instead.
Some people would fix this after the fact (to decide that 10 should be 9). Multiplying it by 2 should give around a ~50% chance of 0 or 1 but will also have a ~0.000000000465% chance of returning a 2 like in Bender's dream.
Saying 0 to 1 as a float might be a bit like mistakenly saying 0 to 10 instead of 0 to 9 as ints when you want ten values starting at zero. In this case because of the broad range of possible float values then it's more like accidentally saying 0 to 1000000000 instead of 0 to 999999999.
With 64bit it's exceedingly rare to overflow but in this case some random functions are 32bit internally so it's not no implausible for that one in two and a half billion chance to occur.
The standard solutions would instead want to be like this:
mt_rand() / (getrandmax() + 1)
There can also be small usually insignificant differences in distribution, for example between 0 to 9 then you might find 0 is slightly more likely than 9 due to precision but this will typically be in the billionth or so and is not as severe as the above issue because the above issue can produce an invalid unexpected out of bounds figure for a calculation that would otherwise be flawless.
Java's Math.random will also never produce a value of 1. Some of this comes from that it is a mouthful to explain specifically what it does. It returns a value from 0 to less than one. It's Zeno's arrow, it never reaches 1. This isn't something someone would conventionally say. Instead people tend to say between 0 and 1 or from 0 to 1 but those are false.
This is somewhat a source of amusement in bug reports. For example, any PHP code using lcg_value without consideration for this may glitch approximately one in a couple billion times if it holds true to its documentation but that makes it painfully difficult to faithfully reproduce.
This kind of off by one error is one of the common sources of "Just turn it off and on again." issues typically encountered in embedded devices.
Solution for PHP 7. Generates random number in [0,1). i.e. includes 0 and excludes 1.
function random_float() {
return random_int(0, 2**53-1) / (2**53);
}
Thanks to Nommyde in the comments for pointing out my bug.
>>> number_format((2**53-1)/2**53,100)
=> "0.9999999999999998889776975374843459576368331909179687500000000000000000000000000000000000000000000000"
>>> number_format((2**53)/(2**53+1),100)
=> "1.0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000"
Most answers are using mt_rand. However, mt_getrandmax() usually returns only 2147483647. That means you only have 31 bits of information, while a double has a mantissa with 52 bits, which means there is a density of at least 2^53 for the numbers between 0 and 1.
This more complicated approach will get you a finer distribution:
function rand_754_01() {
// Generate 64 random bits (8 bytes)
$entropy = openssl_random_pseudo_bytes(8);
// Create a string of 12 '0' bits and 52 '1' bits.
$x = 0x000FFFFFFFFFFFFF;
$first12 = pack("Q", $x);
// Set the first 12 bits to 0 in the random string.
$y = $entropy & $first12;
// Now set the first 12 bits to be 0[exponent], where exponent is randomly chosen between 1 and 1022.
// Here $e has a probability of 0.5 to be 1022, 0.25 to be 1021, etc.
$e = 1022;
while($e > 1) {
if(mt_rand(0,1) == 0) {
break;
} else {
--$e;
}
}
// Pack the exponent properly (add four '0' bits behind it and 49 more in front)
$z = "\0\0\0\0\0\0" . pack("S", $e << 4);
// Now convert to a double.
return unpack("d", $y | $z)[1];
}
Please note that the above code only works on 64-bit machines with a Litte-Endian byte order and Intel-style IEEE754 representation. (x64-compatible computers will have this). Unfortunately PHP does not allow bit-shifting past int32-sized boundaries, so you have to write a separate function for Big-Endian.
You should replace this line:
$z = "\0\0\0\0\0\0" . pack("S", $e << 4);
with its big-endian counterpart:
$z = pack("S", $e << 4) . "\0\0\0\0\0\0";
The difference is only notable when the function is called a large amount of times: 10^9 or more.
Testing if this works
It should be obvious that the mantissa follows a nice uniform distribution approximation, but it's less obvious that a sum of a large amount of such distributions (each with cumulatively halved chance and amplitude) is uniform.
Running:
function randomNumbers() {
$f = 0.0;
for($i = 0; $i < 1000000; ++$i) {
$f += \math::rand_754_01();
}
echo $f / 1000000;
}
Produces an output of 0.49999928273099 (or a similar number close to 0.5).
I found the answer on PHP.net
<?php
function randomFloat($min = 0, $max = 1) {
return $min + mt_rand() / mt_getrandmax() * ($max - $min);
}
var_dump(randomFloat());
var_dump(randomFloat(2, 20));
?>
float(0.91601131712832)
float(16.511210331931)
So you could do
randomFloat(0,1);
or simple
mt_rand() / mt_getrandmax() * 1;
what about:
echo (float)('0.' . rand(0,99999));
would probably work fine... hope it helps you.