Re: Synonym

[Markus Mützel]

Am 08. April 2020 um 18:07 Uhr schrieb "John W. Eaton":
> On 4/8/20 11:46 AM, "Markus Mützel" wrote:
> 
> > I don't think it is necessary to initialize varargout. It should be 
> > initialized to the correct length. [1]
> > This should be enough:
> > 
> > function varargout = short (varargin)
> >    [varargout{:}] = verylongname (varargin{:});
> > endfunction
> 
> Can you verify that this function actually works in Matlab (see below 
> for simple example)?
> 
> > [1]: https://de.mathworks.com/help/matlab/ref/varargout.html
> 
> The example on this page uses
> 
>    [varargout{1:nargout}] = ...
> 
> which does work in Octave.  But
> 
>    [varargout{:}] = ...
> 
> fails in Octave unless varargout is explicitly initialized.
> 
> And there is another example on that page that shows this:
> 
> function varargout = variableNumInputAndOutput(varargin)
>      disp(['Number of provided inputs: ' num2str(length(varargin))])
>      disp(['Number of requested outputs: ' num2str(nargout)])
> 
> If varargout is initialized when the function is called, then why use 
> length (varargin) instead of nargin and nargout instead of length 
> (varargout)?
> 
> Can you try the following in Matlab to verify that varargout is initialized?
> 
> function varargout = xsvd (varargin)
>    numel (varargin)
>    numel (varargout)
>    [varargout{:}] = svd (varargin{:});
> end
> 
> and then call it with
> 
>    xsvd (rand (2))
>    s = xsvd (rand (2))
>    [u, s, v] = xsvd (rand (2))
> 
> Does that work?  If so, I'll fix Octave to match that behavior.

It does not (tested with Matlab R2019b).
I thought it would be the expected behavior by reading this:
"When the function executes, varargout is a 1-by-N cell array, where N is the 
number of outputs requested after the explicitly declared outputs."

But that clearly seems to be *not* the case. I must be misunderstanding what 
that actually means...

Octave's current behavior is compatible to Matlab's.

Markus