Complex Numbers: better implementation

Question

I've already created a Complex class, with all the mathematical methods necessary. But I've seen at the Internet several attempts to do such classes, and one preferred to make a compiler that accepts complex as a primitive type. And he's right. Creating a class to represent the complex numbers takes too much memory and time only to store two doubles. Using C/C++, complex numbers could be made with structs, and a Complex class would be used to implement the necessary methods, wrapping the complex type, like Double wraps a double.
So, I'd like to suggest a new type of class. This class would act like
an C/C++ struct, being able to store only fields and constructors. I believe that this way would make it waste less memory, and enable faster use of this class. It would work like this:
public value class complex {
  public double real;
  public double imag;
public class Complex {
  public static final complex I;
  I.real = 0; I.imag = 1;
}What do you guys think?

Answer 1

Creating a class to represent the complex
numbers takes too much memory and time only to store two doubles.If the only fields in the Complex object are two doubles, how much more memory does a Complex object take up? The methods would only appear in the Class object, not each Complex instance. Right?
I guess you have one reference on the stack and two doubles on the heap, as opposed to two doubles on the stack and nothing in the heap.
So, I'd like to suggest a new type of class. This
class would act like an C/C++ struct, being able to store only
fields and constructors. I believe that this way would
make it waste less memory, and enable faster use of this class.At this point I wonder whether there'd really be a big memory savings, and whether it would be worth watering down the OOP-ness of the language.
Defining a complex primitive is another issue.
Issues like ordering and equality would seem to make the problem really...complex... Perhaps any choice for those issues would only satisfy a third of the interested parties, and any design would satisfy no one.