Should the docstring for a pointers indicate that None is allowed?

[watsonjj]

Similar to sizmailov/pybind11-stubgen#181, I am trying to indicate in the type hints of my pybind11-wrapped code that a value of None is permitted, as this will be converted to nullptr by pybind11.

Currently the pybind11 docstring of a pointer or smart pointer is simply the contained type.

To bypass the current default name for a shared_ptr<Foo>, I can use the following:

// Redefine the docstring signature to account for None (nullptr) input
namespace pybind11 { namespace detail {
template<>
struct type_caster<std::shared_ptr<Foo>> : public copyable_holder_caster<Foo, std::shared_ptr<Foo>> {
public:
    using base = type_caster_base<Foo>;
    static constexpr auto name = const_name("Optional[") + base::name + const_name("]");
};
}}

1. Is there a trick I am missing here? I can't find a simpler way to override the generated docstring
2. Would a PR which changes the name for type_caster<std::shared_ptr<T>> and type_caster<std::unique_ptr<T>> in cast.h be welcome?
3. I am not sure where the name can be overridden for raw pointers...

[VelocityRa]

Agreed, people have been requesting this for years (eg. #3175, #3377).

Pointers in C/C++ can be null. Seems rather a basic aspect to support.
In C++ people usually rely on that and even intentionally choose pointers over references for that purpose.
Why don't pybind types reflect this, even as an option?