Library for PHP to execute functions asynchronously
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Joop Schilder eef63af0b0 Add brackets, update README, refactored naming 3 years ago
lib Add brackets, update README, refactored naming 3 years ago
src Add brackets, update README, refactored naming 3 years ago
.gitignore Add brackets, update README, refactored naming 3 years ago Add brackets, update README, refactored naming 3 years ago
composer.json Add brackets, update README, refactored naming 3 years ago
composer.lock Add brackets, update README, refactored naming 3 years ago



This package is available on Packagist and can be installed using Composer:

$ composer require joopschilder/async-php

It's also possible to manually add it to your composer.json:

    "require": {
        "joopschilder/php-async": "~1.0"

What is this?

This package provides functions to run callables asynchronously in PHP. Return values are shared via System-V shared memory.
To use this package, you need PHP >= 7 with ext-sysvshm and ext-pcntl.

You should consider the state of this package to be highly experimental.

Note: This package should not be used in a CGI environment.
The key that is used to access the block of shared memory is created based on the inode information of one of the source files.
This means that, whenever multiple instances (processes) from the same project source are created, they will try to use the same block of memory and collisions will occur. I might swap the ftok() call for a multi-instance supporting mechanism later on (feel free to do so yourself).

Note: It is possible (but discouraged) to change the amount of available shared memory.
If you wish to do so, it's as simple as calling either Runtime::setSharedMemorySizeMB(<amount of MB>) or Runtime::setSharedMemorySizeB(<amount of bytes>).
If you want to use 32MB for example, call Runtime::setSharedMemorySizeMB(32).
Be sure to make this call before using any of the asynchronous functionalities.

What is this not?

This is, as you probably guessed by now, not intended for use in a production environment.
I'm not saying you can't, I'm just saying you shouldn't.

The code is not unit tested. It has been documented throughout though, so feel free to take a look.



The library exposes three functions in the global namespace that provide indirect access to the Asynchronous class:

  • async(callable $function, ...$parameters) to run something asynchronously, returning a Promise;
  • async_wait_all() to wait for all currently running jobs to finish;
  • async_cleanup() to clean up any zombie processes during runtime if any exist;

A Promise, you say?

async(...) returns an instance of JoopSchilder\Asynchronous\Promise.
A Promise is considered to be resolved when the function it belongs to returned a value or finished execution.
To block execution until a promise is resolved, simply call the resolve() method on the promise.
It's possible to check whether the promise has been resolved in a non-blocking way by calling the isResolved() method.
You can actually return anything that is serializable in PHP: objects, arrays, strings, you name it.

$promise = async(function() {
    sleep(random_int(1, 5));
    return getmypid();

// ... do some other work

$pid = $promise->getValue();

The shutdown handler and destructors should take care of the cleanup.

Asynchronous curl requests

The only reason curl is used here is to provide an intuitive example.
If you really wanted to perform concurrent http requests you should look into either curl_multi_init or just use Guzzle.

$job = function(string $url) {
    $handle = curl_init($url);
    curl_setopt($handle, CURLOPT_FOLLOWLOCATION, 1);
    curl_setopt($handle, CURLOPT_RETURNTRANSFER, 1);
    $response = curl_exec($handle);
    file_put_contents(uniqid('download_'), $response);

$urls = ['', '', 'some.other.domain'];
foreach($urls as $url) {
    async($job, $url);

That's all there is to it.


If you're using a UNIX system, you can make use of the tools ipcs and ipcrm to monitor and manage System V shared memory blocks.
To see what's happening, you can use:

$ watch -n 1 "ipcs -m --human && ipcs -m -p && ipcs -m -t && ipcs -m -u"

To clean all 'unused' shared memory blocks (they might stay in RAM if your program terminated unexpectedly), run

$ ipcrm -a

What's next?

Who really knows?