Scout APM helps PHP developers pinpoint N+1 queries, memory leaks & more so you can troubleshoot fast & get back to coding faster. Start your free 14-day trial today.

Asymmetric visions

There's a new RFC in town called asymmetric visibility, its aim is to define properties which can be written to from a protected or private scope, but which from the outside — the public scope — can only be read.

It looks something like this:

final class Post
{
    public function __construct(
        public private(set) string $title,
    ) {}
}

There are a couple of things going on here:

In other words, this is allowed:

final class Post
{
    public function __construct(
        public private(set) string $title,
    ) {}
    
    public function changeTitle(string $title): void
    {
        // Do a bunch of checks
        if (strlen($title) < 30) {
            throw new InvalidTitle('Title length should be at least 30');
        }
        
        // Change the title from within the private scope
        $this->title = $title;
    }
}

While this isn't:

$post = new Post('Title');

// Setting $title from the public scope isn't allowed:
$post->title = 'Another title';

I would say it's a pretty decent proposal, I can come up with a bunch of use cases where you want public readonly access to an object's properties (without the overhead of implementing getters), while still being able to change a property's value from within its class. That class could for example add internal checks to ensure the value adheres to any number of business rules — as a simplified example: ensuring the title of our post is at least 30 characters long.

So all things good, here's hoping the RFC passes.

Right?

Well, I have a problem with it. Actually, not with the RFC itself — I think it's a very good proposal. No, my concern is not with this RFC on its own, but rather with how it's closely related to readonly properties.

Here we have an RFC which scope overlaps with readonly properties (albeit not entirely the same), with a future possibility to replace readonly properties altogether. Here's a quote from the RFC:

At this time, there are only two possible operations to scope: read and write. In concept, additional operations could be added with their own visibility controls. Possible examples include:

It's clear that this RFC and its successors have the potential to replace readonly properties entirely. Readonly properties — a feature that only has been added one year ago in PHP 8.1, not to mention readonly classes, which are coming to PHP 8.2 later this year.

Despite asymmetric visibility being a great proposal, I'm afraid of what PHP will become if we're adding features only to make them irrelevant three or four years later, as could potentially happen here with readonly properties. We should be very careful and deliberate about how we're changing PHP, and not dismiss existing features too quickly.

If we did, we'd contribute to a lot of uncertainty and instability within the community. Imagine someone adopting readonly properties today, only to hear a year later that by PHP 9.0, they'll probably be deprecated in favor of asymmetric visibility.

Even if readonly properties would stay and coexist with asymmetric visibility, there would be so much room for confusion: when could you use readonly properties? Should you always use asymmetric visibility instead? I would say it's bad language design if a language allows room for these kinds of questions and doubts.

Furthermore, I totally agree with Marco's sentiment on the matter:

I use readonly properties aggressively, and I try to make the state as immutable as possible.

In the extremely rare cases where public get and private set are needed, I rely on traditional getters and setters, which are becoming extremely situational anyway, and still work perfectly fine.

[…]

In fact, I'm writing so few getters and setters these days, that I don't see why I'd need getter and setter semantics to creep into the language, especially mutable ones, not even with the reflection improvements.

Now to be clear: I'm very well aware that asymmetric visibility and readonly properties aren't the same thing. Asymmetric visibility covers a much larger scope and offers much more flexibility. However: Nikita already coined a very similar idea to asymmetric visibility last year, which wasn't pursued in favour of readonly properties. The discussion about whether we want more flexibility has already been had, and the conclusion back then was: no; readonly properties cover 95% of the use cases, and that's good enough.

I would be sad to see PHP become a language that throws out core features every couple of years, for the sake of a little more flexibility. If we wanted more flexibility in this case, we should have decided on that two years ago when readonly properties were discussed in depth; now — in my opinion — is too late.


On a final note, if you are worried about cloning objects with new values (a problem this RFC would solve and readonly properties don't): people are already working on some kind of syntax to allow rewriting readonly properties while cloning. I'd say it's better to focus our efforts in that area, instead of coming up with an entirely different approach.

Even more: the original example I showed with asymmetric visibility allowing for more functionality (internally guarding business rules) wasn't entirely correct. The same is possible with readonly properties, given that we have a way to overwrite readonly values when cloning them:

final class Post
{
    public function __construct(
        public readonly string $title,
    ) {}
    
    public function changeTitle(string $title): self
    {
        // Do a bunch of checks
        if (strlen($title) < 30) {
            throw new InvalidTitle('Title length should be at least 30');
        }
        
        return clone $this with {
            title: $title,
        }
    }
}

Oh and while the above syntax isn't available yet, it's already possible to overwrite readonly properties while cloning today with some additional userland code:

final class Post
{
    use Cloneable;
    
    public function __construct(
        public readonly string $title,
    ) {}
    
    public function changeTitle(string $title): self
    {
        // Do a bunch of checks
        if (strlen($title) < 30) {
            throw new InvalidTitle('Title length should be at least 30');
        }
        
        return $this->with(
            title: $title,
        );
    }
}

In summary: I think asymmetric visibility is a great feature for some use cases, although there are alternatives as well. All in all, I don't think it's worth adding asymmetric visibility now that we have readonly properties. We decided on readonly properties, we'll have to stick with them for the sake of our users and to prevent ambiguous features from making a mess.

I think a unified vision and direction for PHP is lacking these days, and this RFC — great as it is on its own — is a good example of that lacking in practice. I hope that we (PHP internals, that is) can come up with a solution, maybe the PHP Foundation has an important role to play in all of this, in the future?

Noticed a tpyo? You can submit a PR to fix it. If you want to stay up to date about what's happening on this blog, you can follow me on Twitter or subscribe to my newsletter: