Re: Round/curly bracket interchangeability for cell

[Daniel J Sebald]

On 01/09/2017 04:21 PM, Mike Miller wrote:
> On Mon, Jan 09, 2017 at 16:03:30 -0600, Daniel J Sebald wrote:
> > I just noticed that once a cell variable exists, round brackets may be
> > substituted for curly brackets.  I first noticed this when I encountered a
> > simple error like the following:
> >
> > octave:49> a{1} = 10;
> > octave:50> a{2}
> > error: a(2): out of bound 1
> >
> > Notice that the error message has a round bracket rather than curly bracket,
> > in light of the variable 'a' being a cell.  My thinking was that the
> > following makes more sense for an error:
> >
> > error: a{2}: out of bound 1
> >
> > After further examination, I found the following is acceptable:
> >
> > octave:54> b{1} = "curly";
> > octave:55> b(2) = "round"
> > b =
> >
> >
> > {
> >   [1,1] = curly
> >   [1,2] = round
> > }
> >
> > (Well, first thing, notice two spaces between "b =" and the variable
> > contents display, is that a bug?)
> >
> > I don't know if this curly/round bracket is a compatibility issue, but I
> > would think this is too free use of round brackets and can cause more
> > confusion than anything.  I mean, it is difficult enough for someone
> > learning the language to comprehend the difference between a cell and an
> > array.  Then introduce the idea that if I first use curly bracket, from then
> > on I can write lines of code that use the round bracket to index that cell.
> > Being a new user, one might type by mistake "()" rather than "{}" and be
> > completely confused.  Even an experienced user wouldn't like reading code
> > that did such a thing.
>
> Yes, as I recollect this is a compatibility issue.
>
> To further clarify, indexing on cell arrays with round brackets /
> parentheses *is* different from indexing with curly brackets / braces,
> both are intentionally supported and intentionally mean different
> things, as in Matlab:
>
>   >> a = {10, 11, 12};
>   >> a{1}
>   ans =  10
>   >> a(1)
>   ans =
>   {
>     [1,1] = 10
>   }
>
> When used as the LHS of an indexed assignment, they are also different:
>
>   >> a(1:3) = 0
>   a =
>   {
>     [1,1] = 0
>     [1,2] = 0
>     [1,3] = 0
>   }
>   >> a{1:3} = 0
>   error: invalid assignment to cs-list outside multiple assignment
>
>   >> [a{1:3}] = deal (3, 2, 1)
>   a =
>   {
>     [1,1] =  3
>     [1,2] =  2
>     [1,3] =  1
>   }
>
> Only in the degenerate case of indexing a single element of the cell
> array are they equivalent.
>
> Does that clear things up?

Yes, a bit, thanks.  It still might be nice to have the error message 
reflect the bracket usage for the line number that fails, e.g.,

octave:67> [b{1:5}] = deal(a{1:5})
error: a(5): out of bound 4

As pointed out, round bracket and curly bracket behavior is different.

octave:77> clear
octave:78> [a{1:2}] = deal(2)
a =


{
  [1,1] =  2
  [1,2] =  2
}

octave:79> [b{1:2}] = deal(a{1:2})
b =


{
  [1,1] =  2
  [1,2] =  2
}

octave:80> [b{1:2}] = deal((1:2))
b =


{
  [1,1] =

     1   2

  [1,2] =

     1   2

}

Dan