How to get a type instance for Python's int?

[lahwaacz]

In pybind11, the documentation for py::type::of<T>() says:

Convert C++ type to type if previously registered. Does not convert standard types, like int, float. etc. yet. See pybind/pybind11#2486

So I used the py::type::of(handle h) function instead:

[](py::object /* self */) -> py::object {
    return py::type::of( py::int_() );
}

According to the nanobind porting guide, we should replace py::type::of<T>() with nb::type<T>(). However, nb::type does not have an overload with the handle parameter. How can I migrate the code above?

[lahwaacz]

Actually the expression nb::type<int>() compiles, but the resulting object cannot be used in Python:

TypeError: Unable to convert function return value to a Python type! The signature was
    (arg: object, /) -> object

[wjakob]

nb::type<T> is restricted to bindings (e.g., things that are bound with nb::class_<>). This could perhaps be documented better. To directly get the type, use nb::borrow(&PyLong_Type)

[lahwaacz]

Thanks! I have another related question: what is the difference between nb::type<T>() and nb::type_object_t<T>()?

[oremanj]

nb::type<T> is a function; you call it to obtain the Python type object for the nanobind-bound type T. nb::type_object_t<T> is a class; you use it as a parameter type for a bound function, and that parameter will accept any argument that is the type object nb::type<T>() or something derived from it. If you write nb::type_object_t<T>(), you will get a null nb::object; it doesn't look anything up.