Re: Error with pkg.m (== isspace does work with cell arrays anymore)

[Daniel J Sebald]

John W. Eaton wrote:
> On 25-Feb-2008, Søren Hauberg wrote:
>
> | 
> | man, 25 02 2008 kl. 16:17 -0500, skrev John W. Eaton:
> | > On 24-Feb-2008, Michael Goffioul wrote:
> | > 
> | > | I have an error with recent mercurial code, in pkg.m. It reduces to
> | > | the fact that isspace does not work anymore on cell array of
> | > | strings. The error occurs around line 1299:
> | > | 
> | > |   if (! all (isspace (filenames)))
> | > | 
> | > | where filenames is a cell array of strings. I'm not sure what's
> | > | the correct way to fix this.
> | > 
> | > I could probably help if I knew what isspace should do for a cell
> | > array.  Matlab appears to do this:
> | > 
> | >   >> isspace ({' ', ' ', ' '})      
> | > 
> | >   ans =
> | > 
> | >        0     0     0
> | > 
> | > 
> | > Which makes no sense to me.  Is this behavior useful in some way that
> | > I'm not seeing?
> | It seems to be that matlab is doing something like
> | 
> | function output = isspace(input)
> |   if (ischar(input))
> |     ...
> |   else
> |     output = zeros(size(input), "logical");
> |   endif
> | endfunction
> | 
> | As an example:
> |   isspace(@sin)
> | returns 0. Does that make sense? Not in my mind. I think the logic is:
> | 
> |   "A space is part of a string, hence if the input is not a string, it
> | is not a space."
>
> Yes, probably an error is better.
>
> | I don't think we should copy this. If people depend on this behaviour,
> | then there's a pretty good chance of a bug in their code.
>
> OK, isspace appears to be unique.  Matlab doesn't seem to have the
> other is* predicates, but it does have isstrprop, which lumps them all
> in one function.  I propose the following change.  The added functions in
> {Cell,ov-cell}.{h,cc} make writing the isstrprop function fairly
> simple.
>
> I would guess this still doesn't solve the isspace problem in pkg.m.
> We need to know what the intent of the isspace call there is.

I would think that conditional test could be rewritten.  filenames is a cell of 
strings.  It wouldn't be difficult to write a short bit of code to test every 
member of the cell.  One wonders if there were a way to do that with a simple 
change of syntax.  This makes sense:

octave:25> x = [1 2; 3 4]
x =

  1   2
  3   4

octave:26> x(:)
ans =

  1
  3
  2
  4

and this makes sense:

octave:27> filenames = {'this' 'is' 'a' 'test'}
filenames =

{
 [1,1] = this
 [1,2] = is
 [1,3] = a
 [1,4] = test
}

octave:28> filenames{:}
ans =

(,
 [1] = this
 [2] = is
 [3] = a
 [4] = test
,)

And the following makes sense:

octave:30> filenames{1}(:)
ans =

t
h
i
s

so why couldn't we do:

octave:34> filenames{:}(:)
error: can't perform indexing operations for cs-list type
octave:34>

in the instance where every element of the list is a character?

Dan