Property Type in constructor vs dependent

I am using a function handle in a class and there seems to be a difference whether I set it in properties or in dependent properties. In the second case I need to add a cast to my property or else it doesn't work.

There seems to be some type related thing I do not understand.

Thank you for your help!

Function

This is the function and it's execution (in the class D_b is defined as a double):

D_b = 19;
f = @(t)(D_b/2)*(cos(t)+t.*sin(t));
x =  f(linspace(0,pi/4,50)');

Works without cast

When I define it in properties and initialize it via the constructor like below, it works:

properties(SetAccess=immutable, GetAccess=public)
    % Base circle diameter
    D_b double
    % Involute function x axis
    i_x function_handle
end

function obj = Spur()
%....
    obj.D_b = 38; %Example
    obj.i_x = @(t)(D_b/2)*(cos(t)+t.*sin(t));
%....
end

Fails without cast

When I try to use dependents it fails unless I cast D_b/2 to a double:

properties (Dependent)
    D_b double
    i_x function_handle
end
methods
    function db = get.D_b(obj)
        db = double(obj.D*cosd(obj.phiDegrees)); %CASTING
    end
    function ix = get.i_x(obj)
        ix = @(t)(obj.D_b/2)*(cos(t)+t.*sin(t));
    end
end

If I don't cast the following error occures (which is not the case with my first approach):

Error using  * 
Integers can only be combined with integers of the same class, or scalar doubles.

Error in Spur>@(t)(obj.D_b/2)*(cos(t)+t.*sin(t)) (line 97)
ix = @(t)(obj.D_b/2)*(cos(t)+t.*sin(t));

[Mario Malic]

It is a little bit complicated to comment on your code as you are speaking about two issues (or more): one is about casting the output of multiplication of two object properties (which I don't see in your class), and the other one is the method Spur, which apparently works, but, does it? I would suggest to go over OOP basics in MATLAB, there is Onramp and video tutorial by Loren Shure available

Note that my OOP knowledge is a bit limited, so I didn't put this into answer section, but if it helps you, someone will move it.

So, you have error related to this method:

function obj = Spur()
%....
    obj.D_b = 38; %Example
    obj.i_x = @(t)(D_b/2)*(cos(t)+t.*sin(t));
%....
end

What does this do? If you are trying to modify the value of properties of your class, this will not do it, since you haven't supplied the object (instance of class) to the method. This creates a new structure, with fields D_b, and i_x and returns the structure as an output.

If you supplied the object to it, you would probably get an error, because you have set your class properties to be immutable. This means, that properties are set while object is constructed and afterwards, they cannot be changed.

About the casting issue

function db = get.D_b(obj)
    db = double(obj.D*cosd(obj.phiDegrees)); %CASTING
end

I see no problems here, however, I don't see these properties mentioned in your class, so I assume this is different class whose information is missing. You can also inspect values of these object properties by adding the lines above the line with issue or debugging, maybe you forgot to initialize their values.

[Adam]

Spur() is the constructor so this kind of property initialisation works fine. If obj hasn't yet been referenced when you assign to a property in the constructor it will create it at that moment.

For the rest, I'm a little confused what the actual problem is, what is working and what isn't, from the example given.

[TK]

Sorry for being confusing. I will provide both classes below. I wanted to know why in one case my function (i_x) works without casting and in the other case gives me an error when not casting. Where in my code am I loosing the type class information?

Thank you!!!

Works

classdef SpurWorkingWithoutCast
    properties(SetAccess=immutable, GetAccess=public)
        m double
        N uint8 {mustBeInteger, mustBeNonnegative, mustBeGreaterThan(N,1)}
        phiDegrees double {mustBeNonnegative}
        
        D double
        D_b double
        i_x function_handle
        
    end
    methods
        function obj = SpurWorkingWithoutCast(m,N,phiDegrees)
            obj.m = m;
            obj.N = N;
            obj.phiDegrees = phiDegrees;
            obj.D = obj.m*obj.N;
            obj.D_b = obj.D*cosd(obj.phiDegrees);
            obj.i_x = @(t)(obj.D_b/2)*(cos(t)+t.*sin(t));
        end
    end
end

Execute

l = linspace(0,pi/4,50)';
sw = SpurWorkingWithoutCast(2,20,20);
x = sw.i_x(l)

Does not work

classdef SpurWorkingOnlyWithCast
    properties(SetAccess=immutable, GetAccess=public)
        m double
        N uint8 {mustBeInteger, mustBeNonnegative, mustBeGreaterThan(N,1)}
        phiDegrees double {mustBeNonnegative}
    end
    properties (Dependent)
        D double
        D_b double
        i_x function_handle
    end
    methods      
        function d = get.D(obj)
            d = obj.m*obj.N; 
        end
        
        function db = get.D_b(obj)
            %db = double(obj.D*cosd(obj.phiDegrees));
            db = obj.D*cosd(obj.phiDegrees);
        end
        
        function ix = get.i_x(obj)
            ix = @(t)(obj.D_b/2)*(cos(t)+t.*sin(t));
        end
    end
    methods
        function obj = SpurWorkingOnlyWithCast(m,N,phiDegrees)
            obj.m = m;
            obj.N = N;
            obj.phiDegrees = phiDegrees;
        end
    end
end

Execute

l = linspace(0,pi/4,50)';
sf = SpurWorkingOnlyWithCast(2,20,20);
x = sf.i_x(l)

Error using  * 
Integers can only be combined with integers of the same class, or scalar doubles.

Error in SpurWorkingOnlyWithCast>@(t)(obj.D_b/2)*(cos(t)+t.*sin(t)) (line 26)
ix = @(t)(obj.D_b/2)*(cos(t)+t.*sin(t));

Answer 1

[Matt J]

Property type specifiers, such as the double specifier in,

properties (Dependent)

D_b double

end

only perform conversions when a value is assigned to the property, not when values are extracted from it. That is true regardless of whether the property is Dependent or not. However, a consequence of D_b being Dependent in your particular design of SpurWorkingOnlyWithCast is that you never assign and store a value to the D_b property, so the double specifier above isn't really doing anything. Only the get.D_b() method is determining the type of the returned value db.

Note that the property type specifier isn't always irrelevant when dealing with Dependent properties. If you had defined a set.D_b() method in SpurWorkingOnlyWithCast, then the double specifier would cause any assignment of the form obj.D_b=val to pre-convert val to double before passing it to set.D_b().

Ultimately, I think the best solution is to get rid of the casting of N to uint8. There is no reason to have that if you are going to be deriving non-integer quantities from N:

properties(SetAccess=immutable, GetAccess=public)

N {mustBeInteger, mustBeNonnegative, mustBeGreaterThan(N,1)}

...

end

[TK]

Thank you very much for your explanation! Removing the uint8 worked.