Looking for scipy.LowLevelCallable alternative

[matwey]

I am working on Python bindings for my C++ library which implements (among other stuff) something similar to numeric integration. I expect a Python user to provide me with a function callback mapping one float to another. Even than nanobind performance is a way better than pybind11 one currently, I expect that some users would like to use interface similar to one implemented in scipy (scipy.LowLevelCallable) when it is allowed to explicitly pass C-function, Cython function, Numba JIT-compiled or similar to the C side in order to further reduce Python overhead.

How do I do similar in nanobind? pybind11 used to have shortcut for std::function cast, which I never tried though.

[wjakob]

I think that you will have to write your own code that packages up a C function inside a capsule. This is functionality that could be realized on top of a nanobind-based binding, but nanobind's nb_func will not provide such functionality by itself. This is a pretty advanced use case and seems beyond what I can support here.

[matwey]

Got it.