Passing arguments to varargin

Hi

Aesthetic detail:

I have a function using several parameters. These parameters have default values, but can be given as input if desired. I am doing this using varargin in the following format:

function y = my_fun(a,b, varargin)

% Default values
length = 0;
height = 2;

% Loading optional arguments
while ~isempty(varargin)
    switch lower(varargin{1})

          case 'length'
              length = varargin{2};

          case 'height'
              height = varargin{2};

          otherwise
              error(['Unexpected option: ' varargin{1}])
      end

      varargin(1:2) = [];
  end

% the actual function
y = a + b + length*height;

end

This works fine if I write the options directly into the function, e.g. my_fun(a,b,'length',2). What I would like to do is something like this:

options = {'length',1,'height',2};
my_fun(3,4,options);

I know that I can make it work by passing

options{:}

instead of just options, but is there a way to avoid that?

Thanks!

Answer 1

[Adam]

Picking up from your idea you should just be able to use

varargin{:}

inside your function after passing in options rather than passing in options{:} to achieve the same effect.

Personally I tend to do this kind of thing using object-oriented programming, but that is just preference as I do most of my Matlab programming OOP now, so just throwing it out there as an alternative that you might find neater or are free to completely ignore :)

e.g. I have loads of what I call Args classes which are simple classes with all public properties, just like a struct really, but properly defined.

As an example (bear in mind I have just typed this straight off the top of my head not in Matlab so apologies if it contains syntax errors, it is mostly to put the idea across).

classdef MyFuncArgs

     properties( Access = public )

         % All your arguments here with their default values
         varA = 7;
         varB = [15 20 23];

     end

     methods
         function obj = set.varA( obj, varA )
             validateattributes( varA, { 'numeric' }, { 'scalar', 'positive' } );

             obj.varA = varA;
         end

         function obj = set.varB( obj, varB )
             validateattributes( varA, { 'double' }, { 'vector', 'numel', 3, '>' 5, '<=' 25' } );

             obj.varB = varB;
         end
     end
  end

then my function can just take an object of that class as input, I can do a single validateattributes on it to check it is an object of type MyFuncArgs and all the validation of the parameters was done by the Args class so I know in my function that they are there ready for use.

The default values are programmed into the Args class as shown above in the example so that you can either just pass in a default Args class (or create one inside your function if Args is an optional argument to the function) or you can change only as many parameters as you want from defaults and the Args class set functions will ensure you can only set valid parameters.

Where possible I like to set up these Args classes so that the default values are what I most commonly use. So I have a PlotArgs which I use for a Plotter class I have which sets a colourmap, does interpolation, etc, etc in the manner I usually want it, but I can switch off interpolation, change colourmap, etc in that PlotArgs object.

So for example.

args = MyFuncArgs;
args.varB = [10, 12, 13];

y = my_fun( a, b, args );

Answer 2

[Sean de Wolski]

Have you tried inputParser?

It provides a good framework for what you're trying to do.

http://www.mathworks.com/help/releases/R2014b/matlab/ref/inputparser-class.html

Answer 3

[Ced]

Thanks for your help, I wish I could accept both answers.

The inputparser looks really useful if each input needs to fulfil a certain criterion, but overkill for this method.

I have never worked with the OOP framework in matlab, but maybe that would be a good time to start, seems like a very elegant solution...

The solution I've used now is to simply add

% Checking varargin structure
if ( length(varargin) == 1 )
    varargin = varargin{:};
end

at the beginning. If the arguments are passed directly, e.g.

my_fun(a,b,'length',2)

then the if clause is false and nothing happens. Otherwise, a cell was probably passed and the contents are extracted. The correct placement for the inputparser would probably be here, since it would allow to check the content of varargin.

[Adam]

Certainly it would provide a good, relatively simple, route into OOP for a first time, but if the solution you have works and you are pressed for time then just run with that.

You can also use a standard struct in place of a class to define the default arguments, it just isn't as neat as a struct has dynamic fields and you can't create a default struct without a function wrapper or similar.