Re: [classdef] Can someone test this for me?

[Ben Abbott]

On Oct 15, 2013, at 10:11 PM, Michael Goffioul wrote:

> On Tue, Oct 15, 2013 at 10:04 PM, Ben Abbott <address@hidden> wrote:
> 
> On Oct 15, 2013, at 8:53 PM, Michael Goffioul wrote:
> 
> > On Tue, Oct 15, 2013 at 6:38 PM, Michael Goffioul <address@hidden> wrote:
> > On Tue, Oct 15, 2013 at 2:58 PM, Michael Goffioul <address@hidden> wrote:
> > On Tue, Oct 15, 2013 at 2:06 PM, Серёжа Плотников <address@hidden> wrote:
> > New ML Session:
> >
> > >> x = classderiv
> > x =
> >   classderiv with no properties.
> > >> x.fun
> > ans =
> >      0
> > >> addpath a
> > >> x.fun
> > ans =
> >      0
> > >> y = classderiv
> > y =
> >   classderiv with no properties.
> > >> y.fun
> > ans =
> >      0
> >
> > Thank you. I appreciate your help. Those results are not exactly those I 
> > expected, so I'll ask you to run a few more tests.
> >
> > Can you re-run the session above, but start with addpath, that is:
> >
> > addpath a
> > x = classderiv
> > x.fun
> >
> > Then edit classbase.m and remove the "Sealed" attribute and run the 2 
> > sessions (make sure to restart Matlab between each session);
> >
> > x = classderiv
> > x.fun
> > addpath a
> > x.fun
> > y = classderiv
> > y.fun
> >
> > addpath a
> > x = classderiv
> > x.fun
> >
> > Depending on the results, I may ask you (or if anyone else wants to do it) 
> > to run some more tests.
> >
> > Can anybody run these tests for me? That would help clearing out a few gray 
> > zones.
> >
> > Carne Draug has kindly run the test on IRC for me. There's still one thing 
> > I'd like to have tested to confirm my suspicion.
> >
> > Consider the following file/directory structure (assuming "a" and "b" are 
> > not automatically in Matlab path):
> >
> > a/classbase.m:
> > classdef classbase
> >   methods
> >     function y = fun (obj)
> >       y = 0;
> >     end
> >   end
> > end
> >
> > a/@classderiv/classderiv.m:
> > classdef classderiv < classbase
> > end
> >
> > a/@classderiv/fun.m:
> > function y = fun (obj)
> >   y = 1;
> > end
> >
> > b/@classderiv/fun2.m:
> > function y = fun2 (obj)
> >   y = 2;
> > end
> >
> > Then at Matlab prompt:
> >
> > addpath b
> > addpath a
> > x = classderiv
> > x.fun
> > x.fun2
> >
> > Thanks,
> > Michael.
> 
> x =
> 
>   classderiv with no properties.
> 
> 
> ans =
> 
>      1
> 
> No appropriate method, property, or field fun2 for class classderiv.
> 
> Thanks, Ben.
> 
> So the bottom-line seems to be that for classdef classes defined in @-folder, 
> Matlab only looks for methods-defined-in-external-files in the @-directory 
> where the classdef m-file is located. This is rather different than old-style 
> classes, where methods can be spread among various @-folders, and the 
> existence of methods is then sensitive to the path.
> 
> Ben, could you also try the other method-call syntax, to check whether the 
> result is the same (I sincerely hope it is...):
> 
> fun(x)
> fun2(x)
> 
> Thanks,
> Michael.

Sure ...

fun(x)

ans =

     1

fun2(x)

Undefined function 'fun2' for input arguments of type 'classderiv'.

Ben