OOP in MatLab -- assigning values to properties?

I am trying to understand how MatLab does OOP in R2013b, so I got started with a simple test class, but it doesn't seem to behave as desired. The test class has 2 properties, A and B, which I keep at default scope (which I believe is public). I used the following class definition, with default values:

classdef classtest

      properties
          A = 0;
          B = uint64(5);
      end

      methods
          function changeA(obj, newA)
              obj.A = newA;
          end
          function changeB(obj, newB)
              obj.B = newB;
          end
      end

end

Then I run the following code, with outputs shown:

>> a=classtest;
>> a

a = 

    classtest with properties:

      A: 0
      B: 5

>> a.changeA(128);
>> a.changeB(256);
>> a

a = 

    classtest with properties:

      A: 0
      B: 5

>>

The values of A and B have not updated based on the method executed, but are instead still their default values.

What am I doing wrong?

Answer 1

[Sean de Wolski]

Your class is not a handle class but rather a value class. Thus when you run:

 a.changeB(256);

A new variable ans in your workspace will exist with the change. If you want to update the value in this a and keep it as a value class, assign the output back to a:

a = changeA(a,128)

Or change your class to a handle class. It really depends on what you plan to do with this class as to whether you want it as a handle or value class:

classdef classtest < handle

http://www.mathworks.com/help/releases/R2013b/matlab/object-oriented-design-with-matlab.html

[Sam Butler]

Thank you! Extending handle worked as desired. I want to track values similarly to this, so it looks like a handle class is exactly what I want. I guess my C++ background makes me think every class should behave like a "handle" class, but I assume there are times when you don't want that behavior. Regardless, it now works by simply extending handle, as I would expect.

My overall goal with the actual class I am creating is to track the value of a memory location, so I can pass it back and forth to mex and don't lose track of the data location in memory. I prefer using OOP so I can use the delete(obj) destructor method to ensure the memory gets freed even if the programmer is "lazy" (or just forgets to call it).

[Sean de Wolski]

That sounds about right!

The only time I use value classes is when I want them to behave, well I guess, as values. A simple (e.g. double or uint8) matrix is a value class because when you pass it into the function, the function cannot change it. So when making a class that will behave like a matrix, I would use a value class.

[Daniel Shub]

For the OP's definition of changeA, a = changeA(a,128) gives an error. It is worth pointing out that the handle/value class issue is flagged by mlint.

@Sam As for value and handle classes, I think the polynomial class example in the documentation highlights a good use case of a value class.