PHP uniqid函数执行缓慢的问题

前段时间某个需求：客户提交简单的表单可以创建一个适应于全终端（PC，Pad，Phone）的刮刮卡活动H5页面，其中涉及到客户可在线生成限额6W奖品码的功能。

因为需要保持每个活动奖品码的唯一，我们先是准备用PHP的uniqid函数来生成UUID（Universally Unique IDentifier，也叫GUID，为全局唯一标识符，是一种由算法生成的唯一标识）来生成。

但当我们用生成1W测试时候，发现生成过些需要几十秒，还不包括插入至数据库的时间，然后用xhprof写了个简单例子进行性能测试

<?php

xhprof_enable(XHPROF_FLAGS_CPU|XHPROF_FLAGS_MEMORY);

function   myfunc(){

for($i=0;$i<10000;$i++){

$data = uniqid();
}
}

myfunc();

$data = xhprof_disable();

print_r($data);

测试结果：

[myfunc==>uniqid] => Array(
[ct] => 10000
[wt] => 39975062
[cpu] => 0
[mu] => 960752
[pmu] => 0
)

竟然需要接近40秒的时间生成，单次执行需要3969微秒，也即是0.003969秒生成。如果用户提交表单的同时生成兑换码，最坏情况下需要4分钟才给用户反应，当然可以用消息队列异步生成，但是为啥uniqid需要如此多的时间来生成一个简单的字符串呢？

然后去查看uniqid的实现源码，代码贴在下面

PHP_FUNCTION(uniqid)
{
char *prefix = "";
#if defined(__CYGWIN__)
zend_bool more_entropy = 1;
#else
zend_bool more_entropy = 0;
#endif
char *uniqid;
int sec, usec, prefix_len = 0;
struct timeval tv;

if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "|sb", &prefix, &prefix_len,
&more_entropy)) {
return;
}

#if HAVE_USLEEP && !defined(PHP_WIN32)
if (!more_entropy) {
#if defined(__CYGWIN__)
php_error_docref(NULL TSRMLS_CC, E_WARNING, "You must use 'more entropy' under CYGWIN");
RETURN_FALSE;
#else
usleep(1);
#endif
}
#endif
gettimeofday((struct timeval *) &tv, (struct timezone *) NULL);
sec = (int) tv.tv_sec;
usec = (int) (tv.tv_usec % 0x100000);

/* The max value usec can have is 0xF423F, so we use only five hex
* digits for usecs.
*/
if (more_entropy) {
spprintf(&uniqid, 0, "%s%08x%05x%.8F", prefix, sec, usec, php_combined_lcg(TSRMLS_C) * 10);
} else {
spprintf(&uniqid, 0, "%s%08x%05x", prefix, sec, usec);
}

RETURN_STRING(uniqid, 0);
}

看逻辑也没有复杂的操作，也是对现在时间秒和微秒进行简单的处理，然后写了个简单的测试

int getUniqid( char * uid) {
int sec, usec;
struct timeval tv;
gettimeofday(( struct timeval *) &tv, ( struct timezone *) NULL);
sec = ( int) tv. tv_sec;
usec = ( int ) (tv.tv_usec % 0x100000);
sprintf(uid, "%08x%05x" , sec, usec);
return 1;
}

执行1W次也就需要2000微秒，这又是为啥？但是我们发现生成的uid中存在大量的重复，这是才注意点原代码中的usleep函数，
加上usleep函数在测试，这次和PHP结果一致也需要将近40秒，usleep在此为了保持每次生成的uid不同。
那问题出现在usleep函数上了，然后在usleep前后加上取间隔时间，代码如下

struct timeval start, end;
gettimeofday(( struct timeval *) &start, ( struct timezone *) NULL);
usleep(1);
gettimeofday(( struct timeval *) &end, ( struct timezone *) NULL);
unsigned long space = (end.tv_sec - start. tv_sec) * 1000000 + end.tv_usec
- start. tv_usec;
spaceCost += space;

最后发现生成1W奖品码需要39.99587739.995877秒，而usleep间隔时间总和39.982442m，通过打印usleep时间发现每次usleep(1)从进程挂起到唤醒需要4000微秒,本来就知道usleep达不到精度，也也相差太远了点。
最后用了下面代码生成奖品码

/**
* 生成兑换码并保存到数据库  返回setNo
* $pageId 活动ID
* $level 奖品等级
* $numbers 生成奖品的个数
*/
public static  function generateCDKEYAndSave($pageId,$level,$numbers){

$level1Prefix =array(2,5,9,'E','F','M','N','Q','K','Z');//一等奖的前缀

$level2Prefix =array(1,3,7,'A','C','J','R','U','V','X');//二等奖的前置

$level3Prefix = array(4,6,8,'B','D','G','H','I','L','O','P','R','S','T','W','Y');//三等奖的前缀

if(empty($pageId) || empty($level) || empty($numbers)) return false;

$levelPrefix =$level1Prefix;

if($level==2) $levelPrefix = $level2Prefix;

if($level==3) $levelPrefix = $level3Prefix;

$codes =array();

$now = time();

for($i=0;$i<$numbers;$i++){

$prefixKey = array_rand($levelPrefix);

$prefix = self::COUPON_PREFIX.$levelPrefix[$prefixKey];

//$code =base_convert(hexdec(md5(uniqid())),10,26); 服务器上面uniqid执行慢的要死

//$code =base_convert(hexdec(md5($pageId.'A#1$v&'.$i)),10,26);//数据过多 hexdec丢失大量精度

$code1 = base_convert(substr(md5($pageId.$i.$now), 0, 10), 16, 36);

$code2 = base_convert($i, 10, 26);

$code2Len = strlen($code2);

if ($code2Len == 1) {
$code2 .= chr(rand(82, 90)) . chr(rand(82, 90)) . chr(rand(82, 90));
}
else if ($code2Len == 2) {
$code2 .= chr(rand(82, 90)) . chr(rand(82, 90));
}
else if ($code2Len == 3) {
$code2 .= chr(rand(82, 90));
}

$code =$code1.$code2;

$codes[] = $prefix.strtoupper($code);

}

return $codes;

}

附带uuid测试代码

#include <stdio.h>
#include <malloc.h>
#include <sys/time.h>
#include <unistd.h>

unsigned long sleepCost = 0;

int getUniqid( char * uid,int times) {

struct timeval start, end;

gettimeofday(( struct timeval *) &start, ( struct timezone *) NULL);

usleep(1);

gettimeofday(( struct timeval *) &end, ( struct timezone *) NULL);

unsigned long space = (end.tv_sec - start. tv_sec) * 1000000 + end.tv_usec
- start. tv_usec;

sleepCost += space;

if (0 == times%1000) printf ("\n-----sleep cost-------\n%lu usec\n", space);

int sec, usec;

struct timeval tv;

gettimeofday(( struct timeval *) &tv, ( struct timezone *) NULL);

sec = ( int) tv. tv_sec;

usec = ( int ) (tv.tv_usec % 0x100000);

sprintf(uid, "%08x%05x" , sec, usec);

return 1;
}

int main( int argc, char * argv[]) {

struct timeval start, end;

gettimeofday(( struct timeval *) &start, ( struct timezone *) NULL);

for ( int i = 1; i <= 10000; i++) {

char data[20];

getUniqid(data,i);

}
gettimeofday(( struct timeval *) &end, ( struct timezone *) NULL);

unsigned long space = (end.tv_sec - start. tv_sec) * 1000000 + end.tv_usec
- start. tv_usec;

printf( "\n-----cost-------\n% lu usec\n  \n-----sum sleep sost-------\n% lu usec\n" , space,sleepCost);

}