It doesn't have to be?
It could just be an inline class expression (without the new keyword) that
evaluates to a string, e.g. a machine-generated class-name?
I don't think that's what first-class support means? e.g. nothing like
constructors in JS or class instances like Dart?
This could just be a "cheap hack" with machine-generated class-names, so we
can reference anonymous classes by name, the same way we reference classes
now.
Introducing classes as instances of some kind of "class class" would be a
more drastic change, since this would not be compatible with existing
string type-hints for class-names. (?)
On Wed, Feb 13, 2019 at 9:16 AM Joe Watkins <krak...@gmail.com> wrote:
> What you describe is first class support for classes, nothing much to do
> with anonymous classes.
>
> On Wed, 13 Feb 2019, 09:01 Rasmus Schultz <ras...@mindplay.dk wrote:
>
>> The fact that the anonymous class syntax defines a class *and* immediately
>> constructs an instance is quite annoying.
>>
>> For one, this is quite incompatible with DI containers' ability to resolve
>> constructor arguments.
>>
>> Lets say you DI container can register a named component by merely
>> referencing a class that uses constructor injection - so lets say this
>> works:
>>
>> class MyController {
>> public function __construct(MyService $service) {
>> // ...
>> }
>> }
>>
>> $container->register("my-controller", MyController::class);
>>
>> Now I want to register an anonymous class, for example as part of an
>> integration test-suite, which is common enough:
>>
>> $container->register("my-controller", new class {
>> public function __construct(MyService $service) {
>> // ...
>> }
>> });
>>
>> This doesn't work, because you're expected to actually pass the
>> constructor
>> arguments immediately - because you can only define an anonymous class
>> while immediately creating an instance.
>>
>> What I really want is just an anonymous class - not an instance, so:
>>
>> $container->register("my-controller", class {
>> public function __construct(MyService $service) {
>> // ...
>> }
>> });
>>
>> The question is, what would a class expression without the new keyword
>> evaluate to?
>>
>> Since we normally reference classes with just a class-name, I guess I'd
>> expect a string, like you'd get from the ::class constant.
>>
>> Any hope for something like that?
>>
>
