Re: structfun: more than one output argument

[Thorsten Meyer]

Jaroslav Hajek wrote:

> I should note that that the iscell (varargout{1}) is not sufficient to
> check for the uniform output case. Consider
> cellfun (@(x) {x}, {1, 2, 3}, "uniformoutput", true) - the result is
> still a cell.
> I pushed a different solution that really checks for the option, and
> also replaced the loop by a second cellfun call
As structfun is defined now, it is still possible to pass additional
cell arguments down to cellfun, e.g.:
structfun(@(x,y) x.^2+y, struct("a", 1, "b", {1,2}), {3},
"uniformoutput", false)

In your patch you have
  uo_str = "uniformoutput";

  nargs = length (varargin);
  if (nargs >= 2 && strcmpi (varargin{1}, uo_str))
    uniform_output =varargin{2};
  elseif (nargs >= 4 && strcmpi (varargin{3}, uo_str))
    uniform_output =varargin{4};
  endif

which will not catch the "uniformoutput" in this case.

Are these additional cell arguments an unwanted feature or should the
above lines in structfun be replaced by something like:

index = strcmpi (varargin, uo_str);
if any(index)
  uniform_output = varargin{find(index)(end)};
endif


regards

Thorsten