Home >Backend Development >PHP Tutorial >Implementation method of PHP multi-task second-level timer
This article mainly introduces the implementation method of PHP multi-task second-level timer. It is very good and has certain reference value. Friends in need can refer to it
Description
When I was deploying crontab in my company recently, I suddenly thought about whether I could use PHP to implement a timer. The granularity could be down to the second level, because the crontab can reach up to the minute level. I also did some research. After a while, there are not many timers implemented in PHP. The Swoole extension implements a millisecond-level timer, which is very efficient, but after all, it is not written in pure PHP code, so I finally considered using PHP to implement a timer. category for learning reference.
Implementation
When implementing the timer code, two extensions that come with the PHP system are used
Pcntl - Multi-process extension:
Mainly allows PHP to open many sub-processes at the same time and process some tasks in parallel.
Spl - SplMinHeap - Small top heap
A small top heap data structure. When implementing a timer, using this structure is very efficient. The time complexity of insertion and deletion is very low. It is O(logN). Timers like libevent also used rbtree before adopting this data structure after version 1.4. If a linked list or fixed array is used, each insertion or deletion may need to be traversed or sorted again. There are still some performance issues.
Flow Description
1. Define the timer structure, what parameters are there.2. Then register them all into our timer class Timer.
3. Call the monitor method of the timer class to start monitoring.4. The monitoring process is an endless while loop, constantly checking whether the top of the time heap has expired. I originally considered looping to check once every second, but later I thought that there is still a problem in looping to check once every second. If it happens to be When we sleep(1), the timer has expired, so we cannot execute it accurately right away, and there may be a risk of delay, so we still use usleep(1000) to see it in milliseconds and hang the process. Reduce CPU load.Code
/*** * Class Timer */ class Timer extends SplMinHeap { /** * 比较根节点和新插入节点大小 * @param mixed $value1 * @param mixed $value2 * @return int */ protected function compare($value1, $value2) { if ($value1['timeout'] > $value2['timeout']) { return -1; } if ($value1['timeout'] < $value2['timeout']) { return 1; } return 0; } /** * 插入节点 * @param mixed $value */ public function insert($value) { $value['timeout'] = time() + $value['expire']; parent::insert($value); } /** * 监听 * @param bool $debug */ public function monitor($debug = false) { while (!$this->isEmpty()) { $this->exec($debug); usleep(1000); } } /** * 执行 * @param $debug */ private function exec($debug) { $hit = 0; $t1 = microtime(true); while (!$this->isEmpty()) { $node = $this->top(); if ($node['timeout'] <= time()) { //出堆或入堆 $node['repeat'] ? $this->insert($this->extract()) : $this->extract(); $hit = 1; //开启子进程 if (pcntl_fork() == 0) { empty($node['action']) ? '' : call_user_func($node['action']); exit(0); } //忽略子进程,子进程退出由系统回收 pcntl_signal(SIGCLD, SIG_IGN); } else { break; } } $t2 = microtime(true); echo ($debug && $hit) ? '时间堆 - 调整耗时: ' . round($t2 - $t1, 3) . "秒\r\n" : ''; } }
Example
$timer = new Timer(); //注册 - 3s - 重复触发 $timer->insert(array('expire' => 3, 'repeat' => true, 'action' => function(){ echo '3秒 - 重复 - hello world' . "\r\n"; })); //注册 - 3s - 重复触发 $timer->insert(array('expire' => 3, 'repeat' => true, 'action' => function(){ echo '3秒 - 重复 - gogo' . "\r\n"; })); //注册 - 6s - 触发一次 $timer->insert(array('expire' => 6, 'repeat' => false, 'action' => function(){ echo '6秒 - 一次 - hello xxxx' . "\r\n"; })); //监听 $timer->monitor(false);
Execution results
We have also tested extreme situations, At the same time, all 1000 timers expire in 1s, and it only takes 0.126s to adjust the time stack. This is no problem, but every time a timer is adjusted, a child process needs to be started. This may be more time-consuming, and it is possible If these 1,000 items cannot be processed in 1 second, it will affect the monitoring to continue to trigger next time. However, if the child process is not started, for example, it can still be processed if executed directly. . . . Of course there must be a better way, but this is the only thing I can think of at the moment.
Related recommendations:php database cache implementation ideas
The above is the detailed content of Implementation method of PHP multi-task second-level timer. For more information, please follow other related articles on the PHP Chinese website!