MATLAB: Want to create subclass of Dataset class in Stats Toolbox: How to pass arguments to dataset constructor in constructor for new class

Question

How to I pass arguments to the constructor spec = spec@dataset(input); I want to have the full range of input options available to the normal dataset class. That is multiple inputs of the form: DS = DATASET(…, {DATA,'name'}, …)

Should I be trying to create a string for the input argument and passing it using the EVAL(s) function?

classdef ramandataset < dataset
  properties
  ExperimentID
  SampleID
  ExciteWavelength
  DateTaken
  Apparatus
  end
    methods
        function spec = ramandataset(data,names)
        %RAMANDATASET Class Constuctor
        % INPUT DATA: spec = ramandataset(data,names);
        % where data is a double matrix of the form   
        % **********

        % k ri   k ri   k ri
        % **********
        % and names is a 1 by n cell array of strings  
        if nargin == 0
            super_args{1} = [0];
            super_args{2} = '';
        else
           for i=1:size(data,2)/2 
                input = [];
                super_args{1} = data(:,2*i-1:2*i);
                super_args{2} = names{i};
                input = {input super_args};
           end %END FOR
        end %END IF
        spec = spec@dataset(input); %MY PROBLEM IS IN THIS LINE I THINK
        end % END CONSTRUCTOR
    end % END METHODS
    methods
% %         OVERLOADING
    end

end % END CLASSDEF

Any advice would be most appreciated.

Answer 1

I'm not sure what exactly you are trying to do but if you want the full constructor capabilities of the dataset class you will need to allow your own class to take in a variable number of input arguments. So you'll need to use "varargin".

A second thing that might hold you up (it held me up) is how to pass a variable size cell array to the dataset constructor. You have to use "{:}" on the input.

Here's a very simple example:

classdef ramandataset < dataset
  properties
    end
      methods
          function spec = ramandataset(varargin)
              spec = spec@dataset(varargin{:});      
          end % END CONSTRUCTOR
      end % END METHODS 
  end

Heres an exmaple to test this:

data1 = randn(10,1);
data2 = randn(10,1);
ds = ramandataset({data1,'data1'},{data2,'data2'})

I'd start with this and add your custom modifications as you see fit. I'd also expect to overload the subsref and subsasgn functions.

It might take some time to get it all right but its been very worthwhile for my own work. I hope this helps a bit.