Bad luck may follow the same formula: what you did wrong times how many people that affected.

Go and Java claimed they can do without C++, so let them suffer :)

It takes time and because we love Python we did it first. Others will come.

A C++ header file is full of garbage^W implementation details and not-C++ users want a stable API definition - that's what .clif file is about. It's an interface declaration for a native user.

NumPy support is a borderline issue for CLIF.

Typical usage I saw is the following: C++ has some class(es) that best represented as NumPy arrays. So because those classes are specific to the project (not generic like std::) they don't fit into CLIF runtime, but the project supply a C++ library with custom conversion functions as described in ext.md that use NumPy C API tells CLIF that those classes are convertible to Python objects (that NumPy objects are).

That all NumPy integration CLIF needs and it has to come from the user.