Custom function with Mulidimensional Arguments

[Jonathan Scarlett]

Hi,

I'm trying to write a function using the Evaluation-Based Representations framework.  The basic structure of the file is standard:

switch class(varargin{1})

case 'double'

    ...

case 'sdpvar'

    varargout{1} = yalmip('define',mfilename,varargin{:});

case 'char'

    ...

I tried calling the code with a 2x2x2 sdpvar, and it didn't match any of the cases.  I then noticed that the class was in fact "ndsdpvar", so I added 

case 'ndsdpvar'

    varargin{:}

    varargout{1} = yalmip('define',mfilename,varargin{:});

However, this gives the error

??? The following error occurred converting from constraint to logical:

Error using ==> logical

Conversion to logical from constraint is not possible.

Error in ==> yalmip at 169

                if this_hash == internal_sdpvarstate.ExtendedMap(i).Hash

Can the code be adapted to handle multi-dimensional SDPVAR objects?

Thanks.

[Johan Löfberg]

I will look into it, but in the mean time, maybe you can reshape your argument before sending it, and then reshape back (assuming element-wise operator)

 

switch class(varargin{1})

case 'double'

...

case 'sdpvar'
    d = size(varargin{1};

    varargin{1} = reshape(varargin{1},[],1);
    y = yalmip('define',mfilename,varargin{:});
    varargout{1} = reshape(y,d);

case 'char'

[Jonathan Scarlett]

I'm happy with the 'reshape' solution, thanks!