Float the idea: Forward declaring inner classes

[xax...@gmail.com]

I've never written anything to this list before and am trying to follow the instructions here: https://isocpp.org/std/submit-a-proposal

I'd like to allow forward declarations of inner classes.  Here is a situation I've run into where it seems that forward declaring the inner classes would allow this to work:

class Outer1; // <== potentially optional
class Outer::Inner; //<== NEW


class Outer2; // <== potentially optional
class Outer2::Inner; //<== NEW


template<class T>
class SomeClass {

// !!error: no type named 'Inner' in 'Outer1' <== clang error in current c++
  typename T::Inner * some_func();
};


class Outer1 : public SomeClass<Outer1> {
  public:
  class Inner {
  };
};


class Outer2 : public SomeClass<Outer2> {
  public:
  class Inner {
  };
};

While there are certainly workarounds, I find this organizational structure to be useful in other languages and would have used it in c++ had it been available.   

Live code: https://godbolt.org/g/A5jd87

The workaround I've found is this, but the code doesn't match my intent nearly as well.  

class Outer1Inner {};
class Outer1 : public SomeClass<Outer1Inner> {};

Thank you for your feedback - either on the actual proposal or on proposal procedure.

--Zac

[T. C.]

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0289r0.pdf

[xax...@gmail.com]

awesome, thank you.

[xax...@gmail.com]

Oh.  Less awesome that it was rejected.

[T. C.]

It was not rejected. According to [1], EWG apparently liked the idea but thought it needs more work.

[1] https://botondballo.wordpress.com/2016/03/21/trip-report-c-standards-meeting-in-jacksonville-february-2016/

[Nicol Bolas]

And FYI: the correct term is "nested classes". The term "inner class" is primarily a Java term that has additional features that C++ nested classes don't have.