Expose allocated memory from C++ to python in trampoline class

[ThoreWietzke]

This thread is the outcom from issue #330.

I have the following function, which allocates memory and calls a member function of a class which modifies the data in the allocated Memory:

void foo(Base* base)
{
    double* data = allocateMemory();
    Eigen::Map<Eigen::VectorXd> mappedData (data, n);
    base->func(mappedData);
    // do something with data
}

The base class looks like

class Base
{
    public:
        ~Base() {};
        virtual void func(Eigen::Ref<Eigen::VectorXd> in) = 0;
};

with a trampoline to Python

class PyBase : public Base 
{
    NB_TRAMPOLINE(Base, 1);
    public:

        double func(Eigen::Ref<Eigen::VectorXd> in) override
        {
            NB_OVERRIDE_PURE(
                func, /*Name of function*/
                in /*arguments*/   
            );
        }
};

I want to expose the allocated memory inside foo() in the corresponding Python implementation of func(). In pybind11 this worked.
Is this the way to go in nanobind? Are there other ways to expose allocated memory to Python?

[wjakob]

This issue is resolved given that #334 is merged.