Re: Matlab-style code in test/classes

[Robert T. Short]

On 01/30/2013 06:30 AM, Ben Abbott wrote:
> On Jan 30, 2013, at 9:16 AM, Jordi Gutiérrez Hermoso wrote:
>
> > On 30 January 2013 02:59, Julien Bect <address@hidden> wrote:
> > > In my humble opinion: it's better if it stays, if you want to maintain
> > > compatibility in the long run, when both Octave and Matlab will keep on
> > > evolving.
> > "Compatibility" is a fuzzy concept. If it means rewriting Octave code
> > to make sure it runs in Matlab, I don't want anything to do with it.
> > To me, "comptability" just means the other way around: if it runs in
> > Matlab, it should run in Octave.
> >
> > - Jordi G. H.
> I dont' think there is any objection for adding tests that only run in Octave.  
> But I do like the idea of having/keeping tests for compatibility (if it runs in 
> Matlab, it should run in Octave).
>
> Ben
I agree.  If we have a feature, e.g. "if it runs in MATLAB then it 
should run in octave" then the tests should support it.  In particular, 
the legacy class tests were built specifically to make sure that the 
changes I was making to octave would run exactly the same way as they do 
in MATLAB.  I wrote scripts, ran them in MATLAB, then ran them in octave 
to make sure things came out the same.  It seems like a shame to go put 
in a bunch of "endif" statements just for the sake of making the tests 
MATLAB incompatible.  I am not aware of any octave-only extensions to 
legacy classes, so why not make them MATLAB friendly.

I realize this is all pretty nebulous since there isn't a good way to 
automate checking compatibility, but I still think that checking for 
MATLAB compatibility shouldn't be an intolerable burden should someone 
need to do so.

Bob