Re: Archetype for C++ function

[Rik]

On 12/07/2015 04:36 PM, John W. Eaton wrote:
> On 12/06/2015 11:07 PM, Rik wrote: >> 12/6/15 >> >> jwe, >> >> The switch to exceptions in the core is making the code a lot clearer, in >> my mind anyways.  I want to propose a new archetype for C++ functions, and >> if you like it, we can possibly have it as one of the code sprint topics. > > While we are at it, what do you think about writing things like > >   retval = ovl (DiagMatrix (nr, nr, 1.0), >                 Matrix (nr, nc), >                 DiagMatrix (nc, nc, 1.0)); > > instead    of > >   retval(2) = DiagMatrix (nc, nc, 1.0); >   retval(1) = Matrix (nr, nc); >   retval(0) = DiagMatrix (nr, nr, 1.0); > 

I'm not too much bothered either way.  The proposed method does have the advantage of not having to either pre-declare the size of the octave_value_list at construction, use resize, or use indexing with the largest value first.

I do think it is clearer that when a function is declared as "function [x, y, z] = funcname (...)" we return the output variables in the same order as the declaration.  In order to do that I think either we should use the ovl() overload or use

octave_value_list retval (3);
... code ...
retval(0) = DiagMatrix (nr, nr, 1.0);
retval(1) = Matrix (nr, nc);
retval(2) = DiagMatrix (nc, nc, 1.0);

--Rik