Define shared subfunctions inside a classdef *.m file?

I want to define a relatively small class (small enough that it makes sense to do everything inside of one file). I want to reuse a common subfunction between several methods, and having this subfunction inside of the classdef would make it cleaner, easily portable and automatically private. However, I have not been able to do so. Is there any way to accomplish it?

>> d = dummy;
>>??? Undefined function or variable 'my_pi'.
>>
>>Error in ==> dummy>dummy.dummy at 10
>>          obj.dummy = my_pi();  

dummy.m:
classdef dummy
  properties 
      value;
  end
  methods
      function y = my_pi()
          y = 3.141592;
      end        
      function obj = dummy()
          obj.dummy = my_pi();
      end
  end
end

Answer 1

[per isakson]

You can define subfunctions in the same file after the definition of the class, ie

   classdef dummy
   ...
   ...
   end % classdef
   function out = subfun1( in )
   ...
   end

/ per

Answer 2

[David Young]

Non-static methods require an object as the first argument, and as my_pi is not declared static and doesn't have an argument, it isn't found.

Rather than putting my_pi at the end of the file, I would include it properly in the class by declaring it to be static:

classdef dummy
    properties
        value;
    end
    methods
        function obj = dummy()
            obj.value = dummy.my_pi();
        end
    end
    methods (Static)       
        function y = my_pi()
            y = 3.141592;
        end
    end
end

You need to use the class name in the call to the static method. I've also corrected the assignment in the constructor: I assume that you intented to set the value property.

You could also set my_pi's access to private if you want it only to be visible from other methods of the class.

[Knut]

Thank you Per and David. I think the latter post contained a better solution.

[David Young]

But you accepted the former post!

[per isakson]

The static method is certainly the solution preferred from an OO point of view.

[Jiro Doke]

@David or @per isakson,

Can you explain why a static method is better here than a subfunction at the end? If this function is private and only used internally, wouldn't it be better to declare it as a subfunction?

[David Young]

Jiro: First, you can set the access of a method to private - so the private access is explicit, not implicit as for a subfunction. Second, if a file starts "classdef" I expect the file to finish with the matching "end" - it seems messy to me to have bits and pieces hanging around after that. Third, it's cleaner and more consistent to stay within the OO paradigm, at least within a single m-file. Third, because the static method requires the class name to call it, you know exactly where it's to be found, and there's no possibility of getting mixed up with another function of the same name. Fourth, you can easily make the method public if you want to by changing its attributes, whereas the subfunction would need to be moved into a separate file. So overall, it's about cleanness and consistency.

[per isakson]

"Better" can mean different things. I second David's description why a static method is better. I might add that programmers (other than yourself) with experience from other/real OO-languages will understand your code in less time. However, I often use the construct with a subfunction because it helps me make the line with the call more readable. The class typically has a long descriptive name, which would dominate and often force me to use continuation lines. This is more true in cases when the subfunction is called several times.

[David Young]

Agreed - it does depend on the programming context. Also, I see I can't count. Oh well...

[Knut]

@David:

Yes, I accepted the first reply before I saw yours, because it did solve my problem, your post just solved it more elegantly (for me). Now, I cant seem to change the acceptance.