optimization toolbox

				function Y=diff(varargin)
				%DIFF (overloaded)
				
				% Author Johan L鰂berg
				% $Id: diff.m,v 1.1 2006/08/10 18:00:19 joloef Exp $
				
				X = varargin{1};
				n = X.dim(1);
				m = X.dim(2);
				
				Y = X;
				x_lmi_variables = X.lmi_variables;
				lmi_variables = [];
				
				% Fast diff of large matrices. Convert to optmized case
				switch nargin
				    case 1
				        if (min([n m]) > 1)
				            Y = diff(X',1,2)';
				            return
				        end
				
				    case 2
				        if varargin{2} == 1 & (min([n m]) > 1)
				            Y = diff(X',1,2)';
				            return
				        end
				    case 3
				        if varargin{2}==1 & varargin{3} == 2 & (min([n m]) > 1)
				            shift = [-speye(m-1) spalloc(m-1,1,0)] + [spalloc(m-1,1,0) speye(m-1)];
				            %            shift = shift(1:m-1,1:m);
				            shift = kron(shift,speye(n));
				            Y.basis = shift*X.basis;
				            Y.dim(1) = n;
				            Y.dim(2) = m - 1;
				            % Reset info about conic terms
				            Y.conicinfo = [0 0];
				            Y = clean(Y);
				            return
				        elseif varargin{2}==1 & varargin{3} == 1  & (min([n m]) > 1)
				            Y = diff(X',1,2)';
				            return
				        end
				
				    otherwise
				        error('To many input arguments.')
				end
				
				% Slow but safe case. Used for higher order diff
				try
				    j = 1;
				    diffX = diff(reshape(X.basis(:,1),n,m),varargin{2:end});
				    Y.basis=diffX(:);
				    for i = 1:length(x_lmi_variables)
				        diffX = diff(reshape(X.basis(:,i+1),n,m),varargin{2:end});
				        if (norm(diffX,inf)>0)
				            Y.basis(:,j+1) = diffX(:);
				            lmi_variables = [lmi_variables x_lmi_variables(i)];
				            j = j+1;
				        end
				    end
				    if j~=1
				        Y.lmi_variables = lmi_variables;
				    else
				        Y = full(reshape(Y.basis(:,1),size(diffX,1),size(diffX,2)));
				        return
				    end
				catch
				    error(lasterr)
				end
				Y.dim(1) = size(diffX,1);
				Y.dim(2) = size(diffX,2);
				% Reset info about conic terms
				Y.conicinfo = [0 0];