A simple and efficient solution for concurrently sending HTTP requests using PSR-18 client implementations.
Peak is a library that enables concurrent request sending using a request pool. It leverages the event loop of AMPHP, ReactPHP or PSL to handle and manage the requests concurrently.
- PHP 8.1 or higher.
- A package that supports non-block I/O using Fibers under the hood (now refer as driver).
You can install the package via composer:
composer require fansipan/peak
Additionally, depending on your choice of driver, these packages may also need to be installed.
composer require amphp/pipeline
composer require azjezz/psl
composer require clue/mq-react react/async
Typical applications would use the PoolFactory
class to create a pool.
use Fansipan\Peak\PoolFactory;
/** @var \Psr\Http\Client\ClientInterface $client */
$pool = PoolFactory::createForClient($client);
It will attempt to create async version of the client using AsyncClientFactory
. The supported clients are Guzzle and Symfony HTTPClient (Psr18Client
).
You can use any PSR-18 client implementations with ReactPHP driver. If an unsupported client is used, it will be replaced with the
Browser
HTTP client (requirereact/http
installed).
The Fansipan\Peak\PoolFactory
provides a configured request pool based on the installed packages, which is suitable for most cases. However, if desired, you can specify a particular implementation if it is available on your platform and/or in your application.
First, you need to create your desired driver:
use Fansipan\Peak\Concurrency\AmpDeferred;
use Fansipan\Peak\Concurrency\PslDeferred;
use Fansipan\Peak\Concurrency\ReactDeferred;
// AMPHP
$defer = new AmpDeferred();
// PSL
$defer = new PslDeferred();
// ReactPHP
$defer = new ReactDeferred();
Then create an asynchronous client, which is essentially a decorator for the PSR-18 client:
use Fansipan\Peak\Client\GuzzleClient;
use Fansipan\Peak\Client\SymfonyClient;
use Fansipan\Peak\ClientPool;
// Guzzle
$asyncClient = new GuzzleClient($defer);
// or using existing Guzzle client
/** @var \GuzzleHttp\ClientInterface $client */
$asyncClient = new GuzzleClient($defer, $client);
// Symfony HTTP Client
$asyncClient = new SymfonyClient($defer);
// or using existing Symfony client
/** @var \Symfony\Contracts\HttpClient\HttpClientInterface $client */
$asyncClient = new SymfonyClient($defer, $client);
$pool = new ClientPool($asyncClient);
The send
method accepts an iterator of PSR-7 requests or closures/invokable class which receive an Psr\Http\Client\ClientInterface
instance.
use Psr\Http\Client\ClientInterface;
use Psr\Http\Message\RequestInterface;
use Psr\Http\Message\ResponseInterface;
// Using array
$responses = $pool->send([
$psr7Request,
fn (ClientInterface $client): ResponseInterface => $client->sendRequest($psr7Request),
]);
var_dump($responses[0]);
var_dump($responses[1]);
// Using generator when you have an indeterminate amount of requests you wish to send
$requests = static function (int $total) {
for ($i = 0; $i < $total; $i++) {
yield $psr7Request;
}
}
$responses = $pool->send($requests(100));
As you can see from the example above, each response instance can be accessed using an index. However, the response order is not guaranteed. If you wish, you can assign names to the requests to easily track the specific requests that have been sent. This allows you to access the corresponding responses by their assigned names.
use Psr\Http\Client\ClientInterface;
use Psr\Http\Message\RequestInterface;
use Psr\Http\Message\ResponseInterface;
$responses = $pool->send([
'first' => $psr7Request,
'second' => fn (ClientInterface $client): ResponseInterface => $client->sendRequest($psr7Request),
]);
// Or using generator
$requests = function (): \Generator {
yield 'first' => $psr7Request;
yield 'second' => fn (ClientInterface $client): ResponseInterface => $client->sendRequest($psr7Request);
};
$responses = $pool->send($requests());
var_dump($responses['first']);
var_dump($responses['second']);
Sending an excessive number of requests may either take up all resources on your side or it may even get you banned by the remote side if it sees an unreasonable number of requests from your side.
As a consequence, it's usually recommended to limit concurrency on the sending side to a reasonable value. It's common to use a rather small limit, as doing more than a dozen of things at once may easily overwhelm the receiving side.
You can use concurrent
method to set the maximum number of requests to send concurrently. The default value is 25
.
$response = $pool
->concurrent(10) // Process up to 10 requests concurrently
->send($requests);
Additional requests that exceed the concurrency limit will automatically be enqueued until one of the pending requests completes.
composer test
Please see CHANGELOG for more information on what has changed recently.
Please see CONTRIBUTING and CODE_OF_CONDUCT for details.
If you discover any security related issues, please email [email protected] instead of using the issue tracker.
The MIT License (MIT). Please see License File for more information.