Partially support first-class callable syntax in constant expressions

[pevdh]

Description

The first-class callable syntax is currently not allowed in constant expressions:

class X {
    const FN = strlen(...);
}

var_dump((X::FN)('test'));
// PHP Fatal error:  Constant expression contains invalid operations

I propose partially supporting the first-class callable syntax in a constant expression. Specifically, I propose that PHP allows these two cases in a constant context:

class X {
    const A = strlen(...);
    const B = SomeClass::someStaticMethod(...);
}

While (still) disallowing expressions that cannot be known at compile time:

class X {
    const C = $closure(...);
    const D = $obj->method(...);
    const E = ($obj->property)(...);
    // etc. See the first-class callables RFC for a complete list
}

Example

Symfony's Choice assertion allows the user to specify a callback. It would be nice if we could use the first-class callable syntax in attributes, but this is currently not allowed:

use Symfony\Component\Validator\Constraints as Assert;

class User {
     #[Assert\Choice(callback: Role::values(...))] // Using this syntax here causes an error
     public string $role;
}

enum Role: string
{
    case Customer = 'customer';
    case Admin = 'admin';
    
    public static function values() {
        return array_map(fn (Role $a) => $a->name, self::cases());
    }
}

Implementation

I hacked together a working patch that supports the strlen(...) syntax in constant expressions. I haven't gotten around to implementing SomeClass::someStaticFunction(...) yet. There are probably some edge cases that I'm missing - I'm not too familiar with the PHP code base.

I'm also not sure if this feature would require an RFC. I would imagine it would, though.

Please let me know what you think.

[cmb69]

Yeah, I think there should be an RFC to discuss and clarify the details. Note that there is also #8959 which is a more general feature request.

[github-actions]

There has not been any recent activity in this feature request. It will automatically be closed in 14 days if no further action is taken. Please see https://github.com/probot/stale#is-closing-stale-issues-really-a-good-idea to understand why we auto-close stale feature requests.

[gnumoksha]

Please, Mr Bot, don't close this :)

[cmb69]

See https://wiki.php.net/rfc/howto, if you're interested in pursuing this.

[mvorisek]

I expect this to be fixed by #9301 no RFC proposed yet

[bwoebi]

@mvorisek Not sure what you are referring to. That PR is about static variables in functions, not constant expressions in general.

[github-actions]

There has not been any recent activity in this feature request. It will automatically be closed in 14 days if no further action is taken. Please see https://github.com/probot/stale#is-closing-stale-issues-really-a-good-idea to understand why we auto-close stale feature requests.

[mvorisek]

Any CTE expression like strlen - https://3v4l.org/08SOl - should be supported in constant expression.

[KapitanOczywisty]

If function would be mistakenly added to CTE, but would be causing errors, you can just remove it in the next patch version and no one will notice. Removing function from constant expression would cause breaks. Then as an user, how do you even keep track which function are available for constant expressions? I could also imagine that this would require significant rework to implement.

[github-actions]

There has not been any recent activity in this feature request. It will automatically be closed in 14 days if no further action is taken. Please see https://github.com/probot/stale#is-closing-stale-issues-really-a-good-idea to understand why we auto-close stale feature requests.

[dynasource]

+1