Class method is 'call by value' function, isn't it?

classdef myClass
    
    properties
        prop1
    end
    
    methods
        
        function this_ = myClass(argin1_, argin2_) % constructor
            this_.prop1 = argin1_ + argin2_;
        end   

        function setter1(argin)
            prop1_ = argin + 1;
        end
        
        function this_ = setter2(this_, argin)
            this_.prop1 = argin + 1;
        end
        
    end
end

Matlab editor give me a error message for setter1 method.

It tells me that first input arg must be instance itself, and must be returned after manipulation.

This implies a class method of Matlab is a cal-by-value function.

I have very large fixed size array data as a property of my class,

and want to update a couple of elements in it many times very quickly.

so, definitely I want to avoid 'call by value' function as my method.

What would be a option for me,

Thank you.

Answer 1

[Prince Kumar]

Hi,

One of the input arguments must be an object or array of objects of the defining class. These methods can compute values based on object data, can overload MATLAB built-in functions, and can call other methods and functions.

Either of the following statements is correct syntax for calling a method, where obj is an object of the class defining the methodName method:

obj.methodName(arg);
methodName(obj,arg);

You are creating a "value" class. When you copy a value object to another variable or pass a value object to a function, MATLAB creates an independent copy of the object and all the data contained by the object. The new object is independent of changes to the original object.

You need to create "handle" class.

Instances of classes that derive from the "handle" class are references to the underlying object data. When you copy a "handle" object, MATLAB copies the handle, but does not copy the data stored in the object properties. The copy refers to the same object as the original handle. If you change a property value on the original object, the copied handle references the same change.

Following is the way to create a "handle" class:

classdef NumHandle < handle
    properties
        Number = 1
    end
end

For more information please refer the following resources:

• Handle and Value Class
• Avoid unnecessary copies of data
• Which kind of class to use?

Hope this helps

[Jaeseok]

Thank you. One more question for confirmation, please.

My understanding of your answer is that; once I define my class as handle class, 'obj' notation used 'in' my class method shall be a handle to the instance. So, 'obj' can be used as similar concept with 'this' notation in C++. Correct?

[Prince Kumar]

Yes

[Jaeseok]

very nice, THANK YOU.

Answer 2

[per isakson]

See Comparison of Handle and Value Classes. Your myClass is a value class.

"It tells me that first input arg must be instance itself" Yes, Matlab requires that the instance is passed explicitly as in setter2. That applies to both handle and value classes.

"What would be a option for me" Use a handle class.