Coroutine

This is version 2x, it breaks version 1x, by namespace of global functions, moving all required CONSTANTS to the dependence package. This version includes the new Fiber implementation, with no need for PHP ext-fibers extension, slated for PHP 8.1. All ext-fibers tests, and examples as they have them, has been implemented here in examples/fiber, tests/fiber and tests/FiberTest.php.

For maximum performance it's advisable to install the cross-platform libuv library, the PHP extension ext-uv. See the online book for a full tutorial overview of it's usage.

All libuv socket/stream/udp/tcp like features will need to reimplemented for Windows, the previous assumption they where broken, that is not the case, issue with PHP ext-uv version of libuv being used, and the assumption how the feature should work. As such current implementation using native stream_select for Windows, will be rework/refactored.

This package next version after 2x release, will require uv-ffi an FFI version of ext-uv of libuv, as such no additional binary library package will need to be downloaded/installed, and some current dependencies will no longer be necessary.

For a fundamental conceptional overview read "Concurrency and PHP in relation to modern programming languages, Python, Go, NodeJS, Rust, Etc", currently in draft form, will be posted on dev.io when completed.

This version also implements PHP ext-parallel extension in a way that uses the child process features of libuv. For a quick performance study between a thread and process see Launching Linux threads and processes with clone.

The implement here follows parallel\Runtime,parallel\Future, parallel\Channel, and Functional API specs as defined, but without the many limitations.

The limitations are overcome by using opis/closure package. All ext-parallel extension tests and examples, have been mainly modified here in examples/parallel, tests/parallel to include this package library instead and remove single array parameter requirement to variadic array. The tests with a somewhat major difference have .1 added to filename, some test are skipped that might require yield for proper usage.

The ext-parallel events and sync API are not implemented, don't see a use case and is already internally part of this coroutine package.

---

Table of Contents

• Introduction
• Functions
• Installation
• Usage
• Development
• Todo
• Credits/References
• Contributing
• License

Introduction

What is Async? Simply avoiding the blocking of the very next instruction. The mechanism of capturing and running later.

What issues the current PHP implementations this package address? Ease of use. Amp and ReactPHP both do and get you asynchronous results. But both require more setup by the user/developer to get a simple line of code to run. There is also Swoole and Hhvm, but neither are standard PHP installations.

When using Amp or ReactPHP and some of the packages based upon them, you must not only manage the Callbacks you provided, but also the returned Promise object, and the Event Loop object. These libraries modeling themselves around old school Javascript, where Javascript nowadays moving towards simple async/await syntax. Which brings up this question.

What does ease of use mean? We can start by reviewing other programming languages implementations. And the fact you already have a working application, that you want various parts to run more responsive, which translates to do more.

Go has it's Goroutine keyword go then your function and statements.

These links of Python, Ruby, Rust, Nim, C#, Elixir, Java, and C++ details keywords async/await combined with a function and statements.

When using they return the actual results you want, no meddling with any objects, no callbacks, and upon calling everything to up that point continues to run.

The thing about each, is the history leading to the feature, the actual behavior, the underlying concepts are the same, and a simple calling structure statement. They all reference in various ways the use of yield or iterators, which this package relies upon and makes the following true.

The only thing you will do to make your code asynchronous, is placing yield within, and prefix calling your code with yield to get actual results you want.

With this package you will have a PHP version of async/await, by just using yield.

There are a few helper functions available to tie everything together. Mainly, away() that's similar to Python's create_task(), that behaves like Google's go() keyword, which is included here as a alias function go().

By using, it immediately returns a number, that can be used with gather(), another Python like function, which also behaves like Google's WaitGroup. This will wait and return the result of a code distant/detached for running in the background.

This package follows a new paradigm Behavioral Programming with the concept of B-threads, functional generators.

The base overall usage of Swoole Coroutine, and FaceBook's Hhvm PHP follows the same outline implementations as others and put forth here.

To illustrate further take this comparison between NodeJS and Python from Intro to Async Concurrency in Python vs. Node.js.

// async_scrape.js (tested with node 11.3)
const sleep = ts => new Promise(resolve => setTimeout(resolve, ts * 1000));

async function fetchUrl(url) {
    console.log(`~ executing fetchUrl(${url})`);
    console.time(`fetchUrl(${url})`);
    await sleep(1 + Math.random() * 4);
    console.timeEnd(`fetchUrl(${url})`);
    return `<em>fake</em> page html for ${url}`;
}

async function analyzeSentiment(html) {
    console.log(`~ analyzeSentiment("${html}")`);
    console.time(`analyzeSentiment("${html}")`);
    await sleep(1 + Math.random() * 4);
    const r = {
        positive: Math.random()
    }
    console.timeEnd(`analyzeSentiment("${html}")`);
    return r;
}

const urls = [
    "https://www.ietf.org/rfc/rfc2616.txt",
    "https://en.wikipedia.org/wiki/Asynchronous_I/O",
]
const extractedData = {}

async function handleUrl(url) {
    const html = await fetchUrl(url);
    extractedData[url] = await analyzeSentiment(html);
}

async function main() {
    console.time('elapsed');
    await Promise.all(urls.map(handleUrl));
    console.timeEnd('elapsed');
}

main()

# async_scrape.py (requires Python 3.7+)
import asyncio, random, time

async def fetch_url(url):
    print(f"~ executing fetch_url({url})")
    t = time.perf_counter()
    await asyncio.sleep(random.randint(1, 5))
    print(f"time of fetch_url({url}): {time.perf_counter() - t:.2f}s")
    return f"<em>fake</em> page html for {url}"

async def analyze_sentiment(html):
    print(f"~ executing analyze_sentiment('{html}')")
    t = time.perf_counter()
    await asyncio.sleep(random.randint(1, 5))
    r = {"positive": random.uniform(0, 1)}
    print(f"time of analyze_sentiment('{html}'): {time.perf_counter() - t:.2f}s")
    return r

urls = [
    "https://www.ietf.org/rfc/rfc2616.txt",
    "https://en.wikipedia.org/wiki/Asynchronous_I/O",
]
extracted_data = {}

async def handle_url(url):
    html = await fetch_url(url)
    extracted_data[url] = await analyze_sentiment(html)

async def main():
    t = time.perf_counter()
    await asyncio.gather(*(handle_url(url) for url in urls))
    print("> extracted data:", extracted_data)
    print(f"time elapsed: {time.perf_counter() - t:.2f}s")

asyncio.run(main())

Using this package as setout, it's the same simplicity:

// This is in the examples folder as "async_scrape.php"
include 'vendor/autoload.php';

function fetch_url($url)
{
  print("~ executing fetch_url($url)" . \EOL);
  \timer_for($url);
  yield \sleep_for(\random_uniform(1, 5));
  print("time of fetch_url($url): " . \timer_for($url) . 's' . \EOL);
  return "<em>fake</em> page html for $url";
};

function analyze_sentiment($html)
{
  print("~ executing analyze_sentiment('$html')" . \EOL);
  \timer_for($html . '.url');
  yield \sleep_for(\random_uniform(1, 5));
  $r = "positive: " . \random_uniform(0, 1);
  print("time of analyze_sentiment('$html'): " . \timer_for($html . '.url') . 's' . \EOL);
  return $r;
};

function handle_url($url)
{
  yield;
  $extracted_data = [];
  $html = yield fetch_url($url);
  $extracted_data[$url] = yield analyze_sentiment($html);
  return yield $extracted_data;
};

function main()
{
  $urls = [
    "https://www.ietf.org/rfc/rfc2616.txt",
    "https://en.wikipedia.org/wiki/Asynchronous_I/O"
  ];
  $urlID = [];

  \timer_for();
  foreach ($urls as $url)
    $urlID[] = yield \away(handle_url($url));

  $result_data = yield \gather($urlID);
  foreach ($result_data as $id => $extracted_data) {
    echo "> extracted data:";
    \print_r($extracted_data);
  }

  print("time elapsed: " . \timer_for() . 's');
}

\coroutine_run(main());

Try recreating this with the other pure PHP async implementations, they would need an rewrite first to come close.

---

A Coroutine here are specially crafted functions that are based on generators, with the use of yield and yield from. When used, they control context, meaning capture/release an application's execution flow.

When yield is placed within an block of code, it indicates to the calling function, that an object will be returned instead, the code is not immediately executed.

This package represent that calling function, an scheduler, similar to an event loop. A coroutine needs to be scheduled to run, and once scheduled coroutines are wrapped in an Task, which are a type of Promise.

A task is an object that represents some work to be done, potentially with a result at the end of it. These tasks are registered with a scheduler that is responsible for running them.

Due to the single-threaded nature of PHP (without extensions anyway), we cannot think of a task as doing a single long-running calculation - this will block the single thread until the task is finished.

Instead, tasks must perform work in small chunks/iterations ('ticks') where possible, passing control back to the scheduler at appropriate points. This is known as cooperative multi-tasking (so called because the tasks must cooperate by yielding control voluntarily).

The scheduler is responsible for 'ticking' the scheduled tasks, with each scheduled task being repeatedly 'ticked' until it is complete. It is up to the scheduler implementation how to do this in a way that allows all scheduled tasks to run.

A task can become complete in one of three ways:

The task reaches successful completion, and optionally produces a result
The task encounters an error and fails
The task is cancelled by calling cancel()

When using this package, and the code you are working on contain yield points, these define points is where a context switch can happen if other tasks are pending, but will not if no other task is pending. This can also be seen as breakpoints/traps, like when using an debugger, when triggered, the debugger steps in, an you can view state and step thought the remainder of your code.

A context switch represents the scheduler yielding the flow of control from one coroutine to the next.

A coroutine here is define as an function/method containing the yield keyword, in which will return generator object.

The generator object that's immediately returned, gives us access to few methods, that allow itself to progress.

So here we have a very special case with Generators in that it being part of the PHP language, and when looked at through the lens of how Promise's work, and that's to not block, just execute line and return. The main idea of being asynchronous.

Promises returns an object, that's placed into an event loop queue. The event loop does the actual executing the callback attached to the object. This is really a manual process, with much code state/overhead to manage. This is called an Reactor pattern of execution, dispatches callbacks synchronously.

The mechanics of an event loop is already present when an a generator is put in motion. I see this as an Proactor pattern. Since the action of yielding is the initiator, begins the process of checking resource availability, performing operations/actions at that moment, and handling/returning completion events, all asynchronously.

Take a read of this post, What are coroutines in C++20?

There are two kinds of coroutines; stackful and stackless.

A stackless coroutine only stores local variables in its state and its location of execution.

A stackful coroutine stores an entire stack (like a thread).

Stackless coroutines can be extremely light weight. The last proposal I read involved basically rewriting your function into something a bit like a lambda; all local variables go into the state of an object, and labels are used to jump to/from the location where the coroutine "produces" intermediate results.

The process of producing a value is called "yield", as coroutines are bit like cooperative multithreading; you are yielding the point of execution back to the caller.

This package performs cooperative scheduling, the basics for multitasking, asynchronous programming.

The steps, that's taking place when an yield is introduced.

1. The function is now an coroutine.
2. The object returned is captured by the scheduler.
3. The scheduler, wraps this captured generator object around an task object.
4. The task object has additional methods and features, it could be seen as promise like.
5. The task is now place into an task queue controlled by the scheduler.
6. You run your function, putting everything in motion. Here you are not starting any event loop. What could be seen as an event loop, is the work being done before or after the task is place into action by the scheduler.
7. Where will this task land/return to? Answer: The same location that called it, there are no callbacks.

Step 1, is implemented in other languages with an specific keyword, async.

Steps 2 to 6, is preformed in other languages with an specific keyword, await.

The terminology/naming used here is more in line with Python's Asyncio and Curio usage. In fact, most of the source code method calls has been change to match theres.

This package should be seen/used as an user-land extension, it's usage of yield has been envisioned from RFC creators.

Functions

Only the functions located here and in the Core.php file should be used. Direct access to object class libraries is discouraged, the names might change, or altogether drop if not listed here. Third party library package development is the exception.

The functions for Network related in Stream.php, File System in Path.php, and Processes in Worker.php, all have been namespaced, so use as follows:

use function Async\Path\{ , };
use function Async\Worker\{ , };
use function Async\Stream\{ , };

/**
 * Returns a random float between two numbers.
 *
 * Works similar to Python's `random.uniform()`
 * @see https://docs.python.org/3/library/random.html#random.uniform
 */
\random_uniform($min, $max);

/**
 * Return the value (in fractional seconds) of a performance counter.
 * Using either `hrtime` or system's `microtime`.
 *
 * The $tag is:
 * - A reference point used to set, to get the difference between the results of consecutive calls.
 * - Will be cleared/unset on the next consecutive call.
 *
 * returns float|void
 *
 * @see https://docs.python.org/3/library/time.html#time.perf_counter
 * @see https://nodejs.org/docs/latest-v11.x/api/console.html#console_console_time_label
 */
\timer_for(string $tag = 'perf_counter');

/**
 * Makes an resolvable function from label name that's callable with `away`
 */
\async(string $labelFunction, $asyncFunction);

/**
 * Wrap the value with `yield`, when placed within this insure that
 * any *function/method* will be `awaitable` and the actual return
 * value is picked up properly by `gather()`.
 */
return \value($value)

/**
 * Add/schedule an `yield`-ing `function/callable/task` for background execution.
 * Will immediately return an `int`, and continue to the next instruction.
 * Returns an task Id
 * - This function needs to be prefixed with `yield`
 *
 * @see https://docs.python.org/3.7/library/asyncio-task.html#asyncio.create_task
 */
yield \away($awaitedFunction, ...$args) ;

/**
 * Performs a clean application exit and shutdown.
 *
 * Provide $skipTask incase called by an Signal Handler. Defaults to the main parent task.
 * - Use `get_task()` to retrieve caller's task id.
 *
 * - This function needs to be prefixed with `yield`
 */
yield \shutdown($skipTask)

/**
 * Wrap the callable with `yield`, this insure the first attempt to execute will behave
 * like a generator function, will switch at least once without actually executing, return object instead.
 * This function is used by `away` not really called directly.
 *
 * @see https://docs.python.org/3.7/library/asyncio-task.html#awaitables
 */
\awaitAble($awaitableFunction, ...$args);

/**
 * Run awaitable objects in the tasks set concurrently and block until the condition specified by race.
 *
 * Controls how the `gather()` function operates.
 * `gather_wait` will behave like **Promise** functions `All`, `Some`, `Any` in JavaScript.
 *
 * - This function needs to be prefixed with `yield`
 *
 * @see https://docs.python.org/3.7/library/asyncio-task.html#waiting-primitives
 */
yield \gather_wait(array $tasks, int $race = 0, bool $exception = true, bool $clear = true)

/**
 * Run awaitable objects in the taskId sequence concurrently.
 * If any awaitable in taskId is a coroutine, it is automatically scheduled as a Task.
 *
 * If all awaitables are completed successfully, the result is an aggregate list of returned values.
 * The order of result values corresponds to the order of awaitables in taskId.
 *
 * The first raised exception is immediately propagated to the task that awaits on gather().
 * Other awaitables in the sequence won't be cancelled and will continue to run.
 * - This function needs to be prefixed with `yield`
 *
 * @see https://docs.python.org/3.7/library/asyncio-task.html#asyncio.gather
 */
yield \gather(...$taskId);

/**
 * Block/sleep for delay seconds.
 * Suspends the calling task, allowing other tasks to run.
 * A result is returned If provided back to the caller
 * - This function needs to be prefixed with `yield`
 */
yield \sleep_for($delay, $result);

/**
 * Creates an communications Channel between coroutines, returns an object
 * Similar to Google Go language - basic, still needs additional functions
 * - This function needs to be prefixed with `yield`
 */
yield \make();

/**
 * Send message to an Channel
 * - This function needs to be prefixed with `yield`
 */
yield \sender($channel, $message, $taskId);

/**
 * Set task as Channel receiver, and wait to receive Channel message
 * Will continue other tasks until so.
 * - This function needs to be prefixed with `yield`
 */
yield \receiver($channel);

/**
 * A goroutine is a function that is capable of running concurrently with other functions.
 * To create a goroutine we use the keyword `go` followed by a function invocation
 * @see https://www.golang-book.com/books/intro/10#section1
 */
yield \go($goFunction, ...$args);

/**
 * Modeled as in `Go` Language.
 *
 * The behavior of defer statements is straightforward and predictable.
 * There are three simple rules:
 * 1. *A deferred function's arguments are evaluated when the defer statement is evaluated.*
 * 2. *Deferred function calls are executed in Last In First Out order after the* surrounding function returns.
 * 3. *Deferred functions can`t modify return values when is type, but can modify content of reference to
 *
 * @see https://golang.org/doc/effective_go.html#defer
 */
\defer(&$previous, $callback)

/**
 * Modeled as in `Go` Language.
 *
 * Regains control of a panicking `task`.
 *
 * Recover is only useful inside `defer()` functions. During normal execution, a call to recover will return nil
 * and have no other effect. If the current `task` is panicking, a call to recover will capture the value given
 * to panic and resume normal execution.
 */
\recover(&$previous, $callback);

/**
 * Modeled as in `Go` Language.
 *
 * An general purpose function for throwing an Coroutine `Exception`,
 * or some abnormal condition needing to keep an `task` stack trace.
 */
\panic($message, $code, $previous);

/**
 * Return the task ID
 * - This function needs to be prefixed with `yield`
 */
yield \get_task();

/**
 * kill/remove an task using task id
 * - This function needs to be prefixed with `yield`
 */
yield \cancel_task($tid);

/**
 * Wait for the callable/task to complete with a timeout.
 * Will continue other tasks until so.
 * - This function needs to be prefixed with `yield`
 */
yield \wait_for($callable, $timeout);

/**
 * Wait on read stream/socket to be ready read from.
 * Will continue other tasks until so.
 * - This function needs to be prefixed with `yield`
 */
yield \read_wait($stream);

/**
 * Wait on write stream/socket to be ready to be written to.
 * Will continue other tasks until so.
 * - This function needs to be prefixed with `yield`
 */
yield \write_wait($stream);

/**
 * Wait on keyboard input.
 * Will continue other tasks until so.
 * Will not block other task on `Linux`, will continue other tasks until `enter` key is pressed,
 * Will block on Windows, once an key is typed/pressed, will continue other tasks `ONLY` if no key is pressed.
 * - This function needs to be prefixed with `yield`
 */
yield \input_wait($size);

/**
 * An PHP Functional Programming Primitive.
 * Return a curryied version of the given function. You can decide if you also
 * want to curry optional parameters or not.
 *
 * @see https://github.com/lstrojny/functional-php/blob/master/docs/functional-php.md#currying
 */
\curry($function, $required);

\coroutine_instance();

\coroutine_clear();

\coroutine_create($coroutine);

/**
 * This function runs the passed coroutine, taking care of managing the scheduler and
 * finalizing asynchronous generators. It should be used as a main entry point for programs, and
 * should ideally only be called once.
 *
 * @see https://docs.python.org/3.7/library/asyncio-task.html#asyncio.run
 */
\coroutine_run($coroutine);

use function Async\Worker\{ add_process, spawn_task, spawn_await };

/**
 * Add/execute a blocking `subprocess` task that runs in parallel.
 * This function will return `int` immediately, use `gather()` to get the result.
 * - This function needs to be prefixed with `yield`
 */
yield \spawn_task($command, $timeout);

/**
 * Add and wait for result of an blocking `I/O` subprocess that runs in parallel.
 * - This function needs to be prefixed with `yield`
 *
 * @see https://docs.python.org/3.7/library/asyncio-subprocess.html#subprocesses
 * @see https://docs.python.org/3.7/library/asyncio-dev.html#running-blocking-code
 */
 yield \spawn_await($callable, $timeout, $display, $channel, $channelTask, $signal, $signalTask);
/**
 * Add and wait for result of an blocking `I/O` subprocess that runs in parallel.
 * This function turns the calling function internal __state/type__ used by `gather()`
 * to **process/paralleled** which is handled differently.
 * - This function needs to be prefixed with `yield`
 *
 * @see https://docs.python.org/3.7/library/asyncio-subprocess.html#subprocesses
 * @see https://docs.python.org/3.7/library/asyncio-dev.html#running-blocking-code
 */
yield \add_process($command, $timeout);

use function Async\Path\file_***Any File System Command***;

/**
 * Executes a blocking system call asynchronously either natively thru `libuv`, `threaded`, or it's `uv_spawn`
 * feature, or in a **child/subprocess** by `proc_open`, if `libuv` is not installed.
 * - This function needs to be prefixed with `yield`
 */
yield \file_***Any File System Command***( ...$arguments);

Installation

composer require symplely/coroutine

This version will use libuv features if available. Do one of the following to install.

For Debian like distributions, Ubuntu...

apt-get install libuv1-dev php-pear php-dev -y

For RedHat like distributions, CentOS...

yum install libuv-devel php-pear php-dev -y

Now have Pecl auto compile, install, and setup.

pecl channel-update pecl.php.net
pecl install uv-beta

For Windows there is good news, native async thru libuv has arrived.

Windows builds for stable PHP versions are available from PECL.

Directly download latest from https://windows.php.net/downloads/pecl/releases/uv/0.2.4/

Extract libuv.dll to sample directory as PHP binary executable, and extract php_uv.dll to ext\ directory.

Enable extension php_sockets.dll and php_uv.dll in php.ini

cd C:\Php
Invoke-WebRequest "https://windows.php.net/downloads/pecl/releases/uv/0.2.4/php_uv-0.2.4-7.2-nts-vc15-x64.zip" -OutFile "php_uv-0.2.4.zip"
#Invoke-WebRequest "https://windows.php.net/downloads/pecl/releases/uv/0.2.4/php_uv-0.2.4-7.3-nts-vc15-x64.zip" -OutFile "php_uv-0.2.4.zip"
#Invoke-WebRequest "https://windows.php.net/downloads/pecl/releases/uv/0.2.4/php_uv-0.2.4-7.4-ts-vc15-x64.zip" -OutFile "php_uv-0.2.4.zip"
7z x -y php_uv-0.2.4.zip libuv.dll php_uv.dll
copy php_uv.dll ext\php_uv.dll
del php_uv.dll
del php_uv-0.2.4.zip
echo extension=php_sockets.dll >> php.ini
echo extension=php_uv.dll >> php.ini

Note: Seems there are issues with PHP ZTS on both Windows and Linux when using uv_spawn.

Usage

In general, any method/function having the yield keyword, will operate as an interruption point, suspend current routine, do something else, then return/resume.

function main() {
    // Your initialization/startup code will need to be enclosed inside an function.
    // This is required for proper operations to start.
}

\coroutine_run(\main());

There after, review as below, the scripts in the examples folder.

/**
 * @see https://docs.python.org/3/library/asyncio-task.html#timeouts
 */
include 'vendor/autoload.php';

function eternity() {
    // Sleep for one hour
    print("\nAll good!\n");
    yield \sleep_for(3600);
    print(' yay!');
}

function keyboard() {
    // will begin outputs of `needName` in 1 second
    print("What's your name: ");
    // Note: I have three Windows systems
    // - Windows 10 using PHP 7.2.18 (cli) (built: Apr 30 2019 23:32:39) ( ZTS MSVC15 (Visual C++ 2017) x64 )
    // - Windows 10 using PHP 7.1.19 (cli) (built: Jun 20 2018 23:37:54) ( NTS MSVC14 (Visual C++ 2015) x86 )
    // - Windows 7 using PHP 7.1.16 (cli) (built: Mar 28 2018 21:15:31) ( ZTS MSVC14 (Visual C++ 2015) x64 )
    // Windows 10 blocks STDIN from the beginning with no key press.
    // Windows 7 does non-blocking STDIN, if no input attempted. only after typing something it blocks.
    return yield \input_wait();
}

function needName() {
    $i = 1;
    yield \sleep_for(1);
    while(true) {
        echo $i;
        yield \sleep_for(0.05);
        $i++;
        if ($i == 15) {
            print(\EOL.'hey! try again: ');
        }
        if ($i == 100) {
            print(\EOL.'hey! try again, one more time: ');
            break;
        }
    }
}

function main() {
    yield \away(\needName());
    echo \EOL.'You typed: '.(yield \keyboard()).\EOL;

    try {
        // Wait for at most 0.5 second
        yield \wait_for(\eternity(), 0.5);
    } catch (\RuntimeException $e) {
        print("\ntimeout!");
        // this script should have exited automatically, since
        // there are no streams open, nor tasks running, this exception killed `eternity` task
        // currently, will continue to run
        // task id 2 is `ioWaiting` task, the scheduler added for listening
        // for stream socket connections
        yield \cancel_task(2);
        // This might just be because `main` is task 1,
        // and still running by the exception throw, need more testing
    }
}

\coroutine_run(\main());

/**
 * @see https://golangbot.com/goroutines/
 * @see https://play.golang.org/p/oltn5nw0w3
 */
include 'vendor/autoload.php';

function numbers() {
    for ($i = 1; $i <= 5; $i++) {
        yield \sleep_for(250 * \MS);
        print(' '.$i);
    }
}

function alphabets() {
    for ($i = 'a'; $i <= 'e'; $i++) {
        yield \sleep_for(400 * \MS);
        print(' '.$i);
    }
}

function main() {
    yield \go(\numbers());
    yield \go(\alphabets());
    yield \sleep_for(3000 * \MS);
    print(" main terminated");
}

\coroutine_run(\main());

Development

/**
 * Template for developing an library package for access
 */
public static function someName($whatever, ...$args)
{
    return new Kernel(
        function(TaskInterface $task, Coroutine $coroutine) use ($whatever, $args){
            // Use/Execute/call some $whatever with ...$args;
            //
            if ($done) {
                // will return $someValue back to the caller
                $task->sendValue($someValue);
                // will return back to the caller, the callback
                $coroutine->schedule($task);
            }
        }
    );
}

// Setup to call
function some_name($whatever, ...$args) {
    return Kernel::someName($whatever, ...$args);
}

// To use
yield \some_name($whatever, ...$args);

Todo

• Add WebSocket support, or really convert/rewrite some package.
• Add Database support, base off my maintenance of ezsql.
• Add more standard examples from other languages, converted over.
• Update docs in reference to similar sections of functionally in Python, Go or any other languages.

Credits/References

Nikita Popov Cooperative multitasking using coroutines (in PHP!). Which this package forks Ditaio, restructuring/rewriting.

Parallel class is a restructured/rewrite of spatie/async. The Parallel class rely upon symplely/spawn as a dependency, used for subprocess management/execution, it uses uv_spawn of libuv for launching processes. The Spawn package has opis/closure as an dependency, used to overcome PHP serialization limitations, and symfony/process as a fallback to proc_open for launching processes, in case libuv the PHP-UV extension is not installed.

Contributing

Contributions are encouraged and welcome; I am always happy to get feedback or pull requests on Github :) Create Github Issues for bugs and new features and comment on the ones you are interested in.

License

The MIT License (MIT). Please see License File for more information.