Re: numel(foo{:}) - feature or bug ?

[Ben Abbott]

On Aug 6, 2011, at 2:38 AM, Sergei Steshenko wrote:

> Thanks for the update.
> 
> I rechecked the thread, and if I am not mistaken, in my case it is:
> 
> "
> octave:5> foo{4} = [1 2 3 4; 5 6 7 8; 9 10 11 12]
> foo =
> {
>  [1,1] = a
>  [1,2] = ab
>  [1,3] = abc
>  [1,4] =
> 
>      1    2    3    4
>      5    6    7    8
>      9   10   11   12
> 
> }
> octave:6> numel(foo{:})
> ans =  72 
> ",
> 
> while in your case it is:
> 
> "
> octave.exe:6> numel(foo{:})
> ans =  96          ## Hmmm.....
> ".
> 
> I.e. for the same input data we have different results (72 <-> 96).
> 
> Again, if I'm not mistaken in all these observations, could you please
> reopen the http://savannah.gnu.org/bugs/?33954 bug ?
> 
> Thanks,
>  Sergei.

You are comparing two different cases. Phil obtained 96 with different inputs. 
You used ...

        foo = {"a", "ab", "abc", [1:4;5:8;9:12]};

And Phil used ...

        foo = {"a", "ab", "abc", rand(4)}; 

In each case, numel is returning ...

        fun = @(x) prod (size (x));
        prod (cellfun (fun, foo(2:end))

In you case 2 * 3 * 12 = 72

In Phil's example 2 * 3 * 16 = 96

>From my reading of the discussion, I'm skeptical that everyone understands 
>that foo{:} is a cs-list, or maybe don't understand cs-lists? Consider the 
>short example below ...

        clear a
        foo = {"a", "ab", "abc", [1:4;5:8;9:12]};
        a(1).args = foo{1};
        a(2).args = foo{2};
        a(3).args = foo{3}
        a(4).args = char(foo{4});
        numel (a.args)
        ans = 72
        numel (foo{:})
        ans = 72
        numel (foo{1}, foo{2}, foo{3}, foo{4})
        ans =  72

Both foo{:} and a.args produce cs-lists (I added the char() function to avoid a 
syntax error).

Ben