Re: `argn' ? … still part of Octave?

[Carnë Draug]

On 5 October 2011 13:28, Ben Abbott <address@hidden> wrote:
> On Oct 5, 2011, at 12:07 AM, John W. Eaton wrote:
>> On  4-Oct-2011, Ben Abbott wrote:
>> | I noticed in the doc string for "whos" that "argn" is mentioned ...
>> |
>> | "Automatic variable.  An automatic variable is one created by the 
>> interpreter, for example `argn'."
>> |
>> | Is "argn" still a part of Octave, and if so, what is its function?
>>
>> It's like inputname, but it can return something even if the argument
>> was an expression.
>
> Thanks. I asked because of the bug reported below.
>
>        https://savannah.gnu.org/bugs/?34478
>
> Apparently, "argn" isn't working for the private functions for @-style 
> objects.

Hi

I found this because I wanted to implement a new class and was trying
to mimic the obj.method type of API (for matlab compatibility). The
idea is to use arg (or inputname) to get the name of the variable in
the caller and change its value with assignin('caller',...)


> Playing a bit, I noticed some unexpected behavior (I was curious how argn 
> compared with inputname())

By the way, since inputname just evals args in the caller, it has the same bug.

> function varargout = fubar (varargin)
>  varargout{1} = argn(1:nargin);
>  for n = 1:nargin
>    varargout{2}(n) = {inputname(n)};
>  endfor
> endfunction
>
> a = 1; b = 2; c = 3; d = 4;
>
> [args, inputs] = fubar(a, b, c, d)
> args =
>
> a
> b
> c
> d
>
> inputs =
> {
>  [1,1] = a
>  [1,2] = b
>  [1,3] = c
>  [1,4] = d
> }
>
> [args, inputs] = fubar(a, b, a*b, c, d)
> args = abacd
> inputs =
> {
>  [1,1] = a
>  [1,2] = b
>  [1,3] =
>  [1,4] = c
>  [1,5] = d
> }

I don't think so. In the second example, inputname behaves correctly
because arguments that don't have an identifier, should return an empty
string. The third argument does not have a corresponding variable on
the caller workspace (a*b) so inputname returns an empty string for
those cases.

Carnë