Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:128935
Content-Type: text/plain;
	charset=utf-8
Precedence: list
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3826.700.81\))
Subject: Re: [PHP-DEV] PHP True Async RFC Stage 4
In-Reply-To: <51a4f8e0-2dbe-444c-aeda-162020c022c5@app.fastmail.com>
Date: Thu, 23 Oct 2025 12:40:37 -0400
Cc: Aaron Piotrowski <aaron@trowski.com>,
 PHP Internals <internals@lists.php.net>
Content-Transfer-Encoding: quoted-printable
Message-ID: <E0E58430-DC60-4A45-BEB3-EE8D8F460305@zort.net>
References: <CAMW7n8DGbFnQzjt3s9CcOBwhp+MUOV+K3U=yY=ULOTdUXaOLHQ@mail.gmail.com>
 <0e4e39d6-9cc9-4970-92e0-2463143b4011@app.fastmail.com>
 <CAMW7n8AgHnO0rBjqBkHNhfvV5PBNxwgHBJ762N3FtVxoh5WnOQ@mail.gmail.com>
 <E8D4A549-79EA-48E9-9D20-3608F357F227@trowski.com>
 <CAMW7n8D0wMNQE0zuMhbHstyZsDw=e7ryPPZERn11UFV7knmR-g@mail.gmail.com>
 <37180d8d-85b4-49a3-a672-334bf4329470@app.fastmail.com>
 <CAMW7n8AFaUByiJ=2frgkAFWNPWHS2wYyHiKfwf8dzT7AYnLjWg@mail.gmail.com>
 <2f8524a7-dea2-4fbf-933a-c538d3706253@app.fastmail.com>
 <CAMW7n8DKB+AWp2sWPi_w8ZgLQ_zS2MPNgP6n1G5127Nkk7EgOA@mail.gmail.com>
 <151800a7-1094-49bc-8e43-c593a74741af@app.fastmail.com>
 <CAMW7n8CWMpUmVB+FHjuYNFMzswbW5Lo19C1D6CTri6HVsddwkQ@mail.gmail.com>
 <c1f712d7-4922-4620-a1c6-e8fca72c9a55@app.fastmail.com>
 <CAMW7n8C5KvtCZnEgh=hx_eGeBv32Gh5ngNa+Ug7dr4TYKRGa_w@mail.gmail.com>
 <51a4f8e0-2dbe-444c-aeda-162020c022c5@app.fastmail.com>
To: Edmond Dantes <edmond.ht@gmail.com>,
 Rob Landers <rob@bottled.codes>
From: jbafford@zort.net (John Bafford)


> On Oct 23, 2025, at 04:18, Rob Landers <rob@bottled.codes> wrote:
>=20
> On Thu, Oct 23, 2025, at 09:59, Edmond Dantes wrote:
>>=20
>>=20
>> I=E2=80=99d like you to have the full information necessary to make =
an
>> informed decision.
>>=20
>> There are two or more channel objects that need to be awaited. How
>> should this be implemented?
>>=20
>> Currently
>> ```php
>> [$result1, $result2] =3D awaitAny($channel1, $channel2);
>> ```
>>=20
> As for which one I'd rather have, I'd rather have the one where my =
examples hold true:
>=20
> assert(awaitAll($a, $b, $c) =3D=3D=3D [await($a), await($b), =
await($c)]);
>=20
> It's clear what it does and that it is just a shorthand. When that =
fails, we have issues.
>=20
> =E2=80=94 Rob


Just to add to this, in Swift, a Task (the analogue to this proposal's =
Coroutine) can emit, exactly once, a value or an error (exception). =
Attempting to read its value (or error) multiple times gets you the same =
value back. This is very handy in memoization patterns.

In this example, multiple readers can call getImage() with the same id =
and get a value, regardless of whether they're the first (which kicks =
off image loading), whether the image loading has completed, or whether =
the image load is still in progress. (Technically, one could omit the =
LoadingBox wrapper entirely, but that would cause the Task to persist =
after it's completed, and that's a more expensive structure than a =
simple enum.)

```swift
private enum LoadingBox<T : Sendable> {
	case inProgress(Task<T, Never>) //Task<Success, Failure>
	case ready(T)
=09
	var value: T {
		get async {
			switch self {
			case .inProgress(let task):
				return await task.value
		=09
			case .ready(let value):
				return value
			}
		}
	}
}

private var imageCache: [ImageId : LoadingBox<Image>] =3D [:]

func getImage(_ id: ImageId) async -> Image {
	if let cache =3D imageCache[id] {
		return await cache.value
	}

	let task =3D Task { ...load the image... }
	imageCache[id] =3D task
=09
	return await task.value
}
```

Ultimately, I would want to be able to write code such as this in PHP =
(with the generics, of course, but I'm not holding my breath).

If coroutines sometimes return single values and sometimes return =
multiple values, it becomes much harder to write a generic async =
cache/memoization system, because you can't indicate via the type system =
that a single value is expected.

-John