[Scilab-Dev] list as output in function call

[Samuel Gougeon]

Le 24/04/2019 à 09:25, Antoine ELIAS a écrit :
> Le 23/04/2019 à 22:50, Samuel Gougeon a écrit :
>> Hello,
>>
>> Le 23/04/2019 à 22:28, Antoine ELIAS a écrit :
>> .../...
>>>
>>> Just a question, specific to ndgrid, why do not update the function 
>>> to return a list in case of nargout == 1 ?
>>
>> You likely mean as a unique explicit argout, instead of the varargout 
>> "implicit" list.
>> Wrapping the result in a list would break a lot of code, and would 
>> make addressing the results more complicated for most frequent cases.
>>
> Absolutely not ;)
> I said that only for this specific case "ndgrid". 

Yes, this is what i understood.

> I think that only break non-working code.

With the current implementation, the following works, with argn(1)==1:

--> X = ndgrid(1:4)
  X  =
    1.   1.   1.   1.
    2.   2.   2.   2.
    3.   3.   3.   3.
    4.   4.   4.   4.

> and varargout is a real list ! ( Do not discriminate ^^ )
Yes, it is build as a real list to feed the return. But is not 
/processed/ as a classic list when returning, since it is assigned in a 
distributive way, while it's not the case for a simple list.
So yes we have to discriminate both, as the test below shows it:
--> function varargout = test()
   >  varargout = list("a","b","c")
   > endfunction

--> [a,b,c] = test()
  c  =
  c

  b  =
  b

  a  =
  a

--> function L = test()
   >  L = list("a","b","c")
   > endfunction

--> [a,b,c] = test()
Wrong number of output arguments.

And, alone, returning a list as the first element in or out of varargout 
would not fix the issue:

--> function varargout = test()
   >  varargout = list(list("a","b","c"))
   > endfunction

--> a = list(,,);
--> a(:) = test()
Unable to insert multiple item in a list.

--> function [r,varargout] = test()
   >     if argn(1)==1
   >         r = list("a","b","c")
   >         varargout = list()
   >     end
   > endfunction
--> a = list(,,);
--> a(:) = test()
Unable to insert multiple item in a list.

For the general discussion, i think that, from a user's point a view, 
when using a container as LHS argument, a distributive assignment would 
be *often* very useful. It might be either through a new *.=* 
"operator", or through the classic "=" one but only when for instance 
the RHS is a container as well,

  * either a list with the right number of elements at depth = 0
  * or a cell or struct array with the same sizes as the cell or struct
    (or struct or cell) on the LHS

AFAIR, this kind of feature is already reported as a wish.

Best regards
Samuel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/dev/attachments/20190424/56d48131/attachment.htm>