Re: empty ranges are matrices ?

[John W. Eaton]

On 29-Mar-2005, Jean-Francois Cardoso <address@hidden> wrote:

| It is not a problem that I cannot work around, of course.  But it may
| appear counterintuitive: I was disturbed at first that an empty object
| cannot be "stacked" freely, in the sense that:
| 
| octave> r = 1:-1 ; s = [ r 1  ]
| s = 1
| octave> r = 1:-1 ; s = [ r ; 1  ]
| error: number of columns must match (1 != 0)
| 
| while I can do 
| 
| octave> vide = zeros(0)  ;
| octave> [ vide 1 ]       ;
| octave> [ vide ; 1 ]     ;
| octave> 
| 
| but I now understand that an empty range should be more naturally 
| represented as [](1x0) than as [](0x0).  Hence the above error.

Yes, I think the behavior is consistent, but it does not seem to be
compatible.  Matlab seems to allow

  r = 1:-1 ; s = [ r ; 1  ]

to succeed without any warning, but it will warn for other mismatches,
for example

  [zeros(10,1), 1]

| But we may have a more serious problem here:
| 
| octave> zeros(0)
| ans = [](0x0)
| octave> zeros([])
| warning: zeros (A): use zeros (size (A)) instead
| ans = 0
| octave> 
| 
| Octave does produce a warning, but nothing more.  
| 
| Do we really want zeros([]) to return a SCALAR rather than an empty
| object?

Probably it should just produce an error now.

jwe



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------