MATLAB: Overloaded subsref function with varargout, nargout is incorrect

MATLABnargout malfunctioningnargout r2011bnargout returns numeloverloaded functions

I have created an overloaded subsref function for a class definition I've created to emulate the table datatype in R2011b. The function works robustly in R2015b, but when I use it in R2011b the value returned for nargout is wrong.

For example, with 0-1 output arguments, nargout returns numel(obj). With >=2 output arguments, the value returned for nargout is correct.

I created a simple function in R2011b which uses varargout to test, but it showed no such problem.

I have also created an overloaded numel function within the same class definition file. All of the methods are defined within the same script. Would this cause a conflict in R2011b to call the wrong numel function when calling nargout?

Thanks for any tips, -Luke

Best Answer

I've checked and confirmed the interference of the overloaded numel function. Commenting it out of the class definition eliminates the error. I opened the numel documentation and found the warnings with regard to creating an overloaded numel function and the subsequent effect on the subsref and subsasgn functions. I will continue investigating, as I need the overloaded numel function. -Luke

Related Solutions

MATLAB: H5L.iterate exception and behavior of native matlab functions() method

I do not know why H5L.iterate needs function handles with a file defined. I suggest you file an enhancement request to remove the requirement.

I can give you a workaround. You will need to make two changes.

Create your function handles from a function not a script.
The instance method as class and static method as class function handles must be created in the form of an anonymous function:

cfhi = @() tstClass.testFuncInstance;      % instance method, as class
cfhs = @() tstClass.testFuncStatic;      % static method, as class

If you look at the function line the parser is doing this for you but because it is generating the code the file is not being filled in. In the case of a script the owning file is the calling function or the base workspace.

Just a note here you can't call an instance method in the form classname.method not sure why you are testing that. My recomended syntax for a call to an instance method for iterating would be:

fh=@(loc_id,attr_name) myIterateMethod(obj, loc_id, attr_name);

MATLAB: Is there an example on overloading the SUBSREF function for the class, by using built-in SUBSREF

You will be able to return a comma-separated list from your class by using the builtin SUBSREF function to overload your object's SUBSREF function.

For example, if you have a class @testobj, then you could implement SUBSREF as follows:

function varargout = subsref(obj, S)
[varargout{1:nargout}] = builtin('subsref',obj,S);

This should just use the built-in structure SUBSREF to access the data. You may want to add special cases, for example if you're indexing your object such as the following:

x(1:2)

Then you need to make sure you convert the varargout elements to class objects, such as the following:

function varargout = subsref(obj, S)
[varargout{1:nargout}] = builtin('subsref',obj,S);
if numel(S) == 1 & S.type=='()'
    for i = 1: nargout,
      varargout{i} = class(varargout{i},'testobj');
end

If you are experiencing problems with SUBSREF not returning the correct number of comma separated list outputs, see Related Solution "When indexing into my MATLAB object and returning a comma separated list, why don't I get the correct number of outputs?" listed below.

Best Answer

Related Solutions

MATLAB: H5L.iterate exception and behavior of native matlab functions() method

MATLAB: Is there an example on overloading the SUBSREF function for the class, by using built-in SUBSREF

Related Question