Generating UNIQUE Random Numbers within a range

Array with range of numbers at random order:

$numbers = range(1, 20);
shuffle($numbers);

Wrapped function:

function UniqueRandomNumbersWithinRange($min, $max, $quantity) {
    $numbers = range($min, $max);
    shuffle($numbers);
    return array_slice($numbers, 0, $quantity);
}

Example:

<?php
print_r( UniqueRandomNumbersWithinRange(0,25,5) );
?>

Result:

 Array
(
    [0] => 14
    [1] => 16
    [2] => 17
    [3] => 20
    [4] => 1
)

If you want to generate 100 numbers that are random, but each number appearing only once, a good way would be to generate an array with the numbers in order, then shuffle it.

Something like this:

$arr = array();

for ($i=1;$i<=101;$i++) {
    $arr[] = $i;
}

shuffle($arr);

print_r($arr);

Output will look something like this:

Array
(
    [0] => 16
    [1] => 93
    [2] => 46
    [3] => 55
    [4] => 18
    [5] => 63
    [6] => 19
    [7] => 91
    [8] => 99
    [9] => 14
    [10] => 45
    [11] => 68
    [12] => 61
    [13] => 86
    [14] => 64
    [15] => 17
    [16] => 27
    [17] => 35
    [18] => 87
    [19] => 10
    [20] => 95
    [21] => 43
    [22] => 51
    [23] => 92
    [24] => 22
    [25] => 58
    [26] => 71
    [27] => 13
    [28] => 66
    [29] => 53
    [30] => 49
    [31] => 78
    [32] => 69
    [33] => 1
    [34] => 42
    [35] => 47
    [36] => 26
    [37] => 76
    [38] => 70
    [39] => 100
    [40] => 57
    [41] => 2
    [42] => 23
    [43] => 15
    [44] => 96
    [45] => 48
    [46] => 29
    [47] => 81
    [48] => 4
    [49] => 33
    [50] => 79
    [51] => 84
    [52] => 80
    [53] => 101
    [54] => 88
    [55] => 90
    [56] => 56
    [57] => 62
    [58] => 65
    [59] => 38
    [60] => 67
    [61] => 74
    [62] => 37
    [63] => 60
    [64] => 21
    [65] => 89
    [66] => 3
    [67] => 32
    [68] => 25
    [69] => 52
    [70] => 50
    [71] => 20
    [72] => 12
    [73] => 7
    [74] => 54
    [75] => 36
    [76] => 28
    [77] => 97
    [78] => 94
    [79] => 41
    [80] => 72
    [81] => 40
    [82] => 83
    [83] => 30
    [84] => 34
    [85] => 39
    [86] => 6
    [87] => 98
    [88] => 8
    [89] => 24
    [90] => 5
    [91] => 11
    [92] => 73
    [93] => 44
    [94] => 85
    [95] => 82
    [96] => 75
    [97] => 31
    [98] => 77
    [99] => 9
    [100] => 59
)

The idea consists to use the keys, when a value is already present in the array keys, the array size stays the same:

function getDistinctRandomNumbers ($nb, $min, $max) {
    if ($max - $min + 1 < $nb)
        return false; // or throw an exception

    $res = array();
    do {
        $res[mt_rand($min, $max)] = 1;
    } while (count($res) !== $nb);
    return array_keys($res); 
}

Pro: This way avoids the use of in_array and doesn't generate a huge array. So, it is fast and preserves a lot of memory.

Cons: when the rate (range/quantity) decreases, the speed decreases too (but stays correct). For a same rate, relative speed increases with the range size.(*)

(*) I understand that fact since there are more free integers to select (in particular for the first steps), but if somebody has the mathematical formula that describes this behaviour, I am interested by, don't hesitate.

Conclusion: The best "general" function seems to be a mix between this function and @Anne function that is more efficient with a little rate. This function should switch between the two ways when a certain quantity is needed and a rate (range/quantity) is reached. So the complexity/time of the test to know that, must be taken in account.

$len = 10;   // total number of numbers
$min = 100;  // minimum
$max = 999;  // maximum
$range = []; // initialize array
foreach (range(0, $len - 1) as $i) {
    while(in_array($num = mt_rand($min, $max), $range));
    $range[] = $num;
}
print_r($range);

---

I was interested to see how the accepted answer stacks up against mine. It's useful to note, a hybrid of both may be advantageous; in fact a function that conditionally uses one or the other depending on certain values:

# The accepted answer
function randRange1($min, $max, $count)
{
    $numbers = range($min, $max);
    shuffle($numbers);
    return array_slice($numbers, 0, $count);
}

# My answer
function randRange2($min, $max, $count)
{
    $range = array();
    while ($i++ < $count) {
        while(in_array($num = mt_rand($min, $max), $range));
        $range[] = $num;
    }
    return $range;
}

echo 'randRange1: small range, high count' . PHP_EOL;
$time = microtime(true);
randRange1(0, 9999, 5000);
echo (microtime(true) - $time) . PHP_EOL . PHP_EOL;

echo 'randRange2: small range, high count' . PHP_EOL;
$time = microtime(true);
randRange2(0, 9999, 5000);
echo (microtime(true) - $time) . PHP_EOL . PHP_EOL;

echo 'randRange1: high range, small count' . PHP_EOL;
$time = microtime(true);
randRange1(0, 999999, 6);
echo (microtime(true) - $time) . PHP_EOL . PHP_EOL;

echo 'randRange2: high range, small count' . PHP_EOL;
$time = microtime(true);
randRange2(0, 999999, 6);
echo (microtime(true) - $time) . PHP_EOL . PHP_EOL;

The results:

randRange1: small range, high count
0.019910097122192

randRange2: small range, high count
1.5043621063232

randRange1: high range, small count
2.4722430706024

randRange2: high range, small count
0.0001051425933837

If you're using a smaller range and a higher count of returned values, the accepted answer is certainly optimal; however as I had expected, larger ranges and smaller counts will take much longer with the accepted answer, as it must store every possible value in range. You even run the risk of blowing PHP's memory cap. A hybrid that evaluates the ratio between range and count, and conditionally chooses the generator would be the best of both worlds.