I have have a function in wrote in C++ that calls some functions in a old lib. This function creates some memory makes the calls and destroys the memory. To optimize this I would create an object that would keep the memory allocated until the object was destroyed. However I’m going to be calling this function from C# and don’t believe I can export a Class, just functions or variables.
My idea is instead this; Think of the DLL as a class and the use local vars inside the scope of the dll to point to memory. Then have a function to create the memory, call the worker functions and another to destroy the memory when done with the DLL.
Is this a good approach? Is there a better way?
I prefer to write a managed wrapper in C++/CLI (formerly Managed C++), as it makes it much easier to explicitly do what you want with managed/unmanaged interoperability on the C++ side, and your C# doesn’t get polluted with P/Invoke style code.
Edit Just noticed your comment “However I’m going to be calling this function from C# and don’t believe I can export a Class, just functions or variables.“
That’s not entirely true – C# can import full classes from an assembly generated from C++/CLI code.