Php radomize

PHP と
乱打舞図
2014/03/15
第5回闇PHP勉強会
do_aki

＠do_aki
http://do-aki.net/

title was born from typo
らんだまいず
乱打舞図
モテカルロ法っぽくない？

randomize
• 無秩序に並び替えること
• 規則性がないこと
• 偏りがないこと

http://nmi.jp/archives/541
(http://www.lauradhamilton.com/random-
lessons-online-poker-exploit)

randomize function in php
shuffle
str_shuffle

give it try (php 5.5.10)
function bench($name, $suffle_func) {
$result = [];
for ($i=0; $i<60000;++$i) {
$a = range(1, 3);
$suffle_func($a);
$r = implode('', $a);
$result[$r]=isset($result[$r])? $result[$r]+1 : 1;
}
print "{$name}¥n";
ksort($result);
print_r($result);
}

shuffle / str_shuffle
bench("shuffle", "shuffle");
bench("str_shuffle", function(&$ary) {
$str = str_shuffle(implode('', $ary));
$ary = str_split($str, 1);
});

result of shuffle
shuffle
Array
(
[123] => 9873
[132] => 10104
[213] => 9999
[231] => 9944
[312] => 9998
[321] => 10082
)

result of str_shuffle
str_shuffle
Array
(
[123] => 9956
[132] => 9928
[213] => 10080
[231] => 9969
[312] => 10149
[321] => 9918
)

No Problem
(in current
version of php)

過去の shuffle は等分散ではなかった

implementation of shuffle(current version)
long n_elems, rnd_idx, n_left;
char temp;
/* The implementation is stolen from array_data_shuffle */
/* Thus the characteristics of the randomization are the same */
n_elems = len;
if (n_elems <= 1) { return; }
n_left = n_elems;
while (--n_left) {
rnd_idx = php_rand(TSRMLS_C);
RAND_RANGE(rnd_idx, 0, n_left, PHP_RAND_MAX);
if (rnd_idx != n_left) {
temp = str[n_left];
str[n_left] = str[rnd_idx];
str[rnd_idx] = temp;
}
}
ext/standard/string.c より抜粋 / shuffle の実装も同じ

implementation of shuffle (php < 4.3.0)
PHP_FUNCTION(shuffle){
zval *array;
if (zend_parse_parameters(1 TSRMLS_CC, "a", &array) == FAILURE){
RETURN_FALSE;
}
if (zend_hash_sort(Z_ARRVAL_PP(&array),
(sort_func_t)php_mergesort, array_data_shuffle, 1 TSRMLS_CC) ==
FAILURE) {
RETURN_FALSE;
}
RETURN_TRUE;
}
}
static int array_data_shuffle(const void *a, const void *b
TSRMLS_DC) {
return (php_rand(TSRMLS_C) % 2) ? 1 : -1;
}

＿人人人人人人人人人人人人＿
＞ランダムマージソート＜
￣Y^Y^Y^Y^Y^Y^Y^Y^Y^Y^Y￣

implementation of str_shuffle (php < 4.3.0)
PHP_FUNCTION(str_shuffle)
{
/* Note : by using current php_string_shuffle for string */
/* with 6 chars (6! permutations) about 2/3 of them are */
/* computed for 6! calls for the function. So it isn't so */
/* unique. The ratio is the same for other lengths. */
char *str;
int i, str_len;
i = 0;
if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "s", &str,
&str_len) == FAILURE) {
RETURN_FALSE;
}
zend_qsort((void *)str, str_len, sizeof(char), php_string_shuffle TSRMLS_CC);
RETURN_STRINGL(str, str_len, 1);
}

implementation of str_shuffle
(php < 4.3.0)
static int php_string_shuffle(const void *a,
const void *b TSRMLS_DC)
{
long rnd;
rnd = php_rand(TSRMLS_C);
if (rnd % 3)
return 1;
else if (rnd % 5)
return 0;
else
return -1;
}

＿人人人人人人人人人人人人人人人＿
＞ランダム(?) クイックソート＜
￣Y^Y^Y^Y^Y^Y^Y^Y^Y^Y^Y^Y^Y^Y￣

emulate old shuffle (php)
bench("old_shuffle", function(&$ary){
usort($ary, function($a, $b) {
return (rand() % 2) ? 1 : -1;
});
});
bench("old_str_shuffle", function(&$ary){
usort($ary, function($a, $b) {
$rnd = rand();
if ($rnd % 3) {
return 1;
} elseif ($rnd % 5) {
return 0;
} else {
return -1;
}
});
});

result
old_shuffle
Array
(
[123] => 22358
[132] => 7627
[213] => 3771
[231] => 3827
[312] => 7487
[321] => 14930
)
old_str_shuffle
Array
(
[123] => 30982
[132] => 4453
[213] => 5994
[231] => 2928
[312] => 8979
[321] => 6664
)

現在の shuffle は、本当に問題ないのか？

rand vs mt_rand
• rand
– libc の random or lrand48 or rand
– environment dependent
• mt_rand
– Mersenne Twister (MT19937)
– implementation dependent

re-imprement of shuffle
(environment independent)
function mt_shuffle(&$ary) {
$n = count($ary);
while(--$n) {
$rnd_idx = mt_rand(0, $n);
if ($rnd_idx != $n) {
$tmp = $ary[$n];
$ary[$n] = $ary[$rnd_idx];
$ary[$rnd_idx] = $tmp;
}
}
}

Conclusion
• 現在の shuffle / str_shuffle は、均等に分配
される
• 異なる環境での再現性が必要なら mt_rand
つかって再実装しよう
• php 4 はオワコン

End Of Slide
Let’s enjoy PHP hack life ;-)

Php radomize

More Related Content

What's hot (19)

More from do_aki (20)

Recently uploaded (20)

Php radomize