RFC: Callable spies

[davedevelopment]

Here's a take on the request in #711

 $spy = spy(function($n) { return $n + 1;});

 array_map($spy, [1, 2]); // [2, 3]

 $spy->shouldHaveBeenCalled();
 $spy->shouldHaveBeenCalled()->twice();
 $spy->shouldHaveBeenCalled()->with(1)->once();
 $spy->shouldHaveBeenCalled()->with(2)->once();

 $spy->shouldHaveBeenCalled()->with(3); // throws...

Todos/considerations

• Better error messages? Ideally we could identify that these are a special case and avoid complicating things showing the internals of ClosureWrapper::__invoke etc
• Make it available for mocking as well? It would mean coming up with some syntax or probably overloading expects and allows somehow. Might get too complicated.
• shouldNotHaveBeenCalledWith(123, 456, 789) ? I think I prefer this for specifying arguments with a not

[davedevelopment]

Also probably worth considering if it's even worth putting this in to mockery. It's simple enough to maintain I think, but the functionality isn't too difficult to roll on your own as you don't really need the code generation etc.

[MarcusSchwarz]

This line should not be necessary. ...$args is an array of all function arguments already.

[davedevelopment]

👍

[robertbasic]

@davedevelopment what's the status with this? Does it require any more work?

[davedevelopment]

@robertbasic no further progress. Error messages are not very friendly at the moment. Happy to merge as an experimental feature.

[tflori]

sorry, I somehow missed this MR.

In the mean time I found a problem we would have with this implementation in laravel: they have a lot of functions where they only allow \Closure. But I guess in this case the user has to write his own closure around the wrapper.

So for me this is good

[davedevelopment]

Merged, will leave undocumented as it's still considered experimental