From help-octave-request at bevo dot che dot wisc dot edu  Tue Feb 10 12:54:02 1998
Subject: Empty matrix question
From: Andy Adler <adlera at njc dot org>
To: help-octave at bevo dot che dot wisc dot edu
Date: Tue, 10 Feb 1998 12:00:15 -0700 (MST)

I'm having a problem with empty matrices.

There's probably an octave preference variable to do this
but I couldn't find it. (So far 90% of my questions to 
help-octave have been answered by a preference variable)

Anyway, here's the sample code

  octave:54> vector=ones(1,5);
  octave:55> numfix=0;
  octave:56> vector(1:numfix)=zeros(numfix,1);
  error: A([]) = X: X must also be an empty matrix
  error: evaluating assignment expression near line 56, column 17

"numfix" is not necessarily zero in the "real" code, but can
be zero if there are no elements to "fix"

The problem seems to be that the matrices are different types
of empty matrix

  octave:58> vector(1:numfix) 
  ans = [](0x0)
  octave:59> zeros(numfix,1) 
  ans = [](0x1)

Because this works if I try

  octave:60> vector(1:numfix)=zeros(0,0);


I realize that I could fix this code using something like

  octave:56> vector(1:numfix)=zeros(numfix, (numfix>=1) );

but this seems to be inelegant.

Is there a way to change it?

( On a side note, The abstract of one of the articles
  referenced by the octave docs talks about the inadequacy
  of the MATLAB empty matrix realization. This must explain
  why Octave is implemented the way it is

  Does anyone feel like explaining to me what the deficiencies are?

  [ Nett, CN.  Haddad, WM.IEEE Trans Automatic Control. 38:771-5. 1993
   A system-theoretic appropriate realization of the empty matrix concept
   Abstract: A system-theoretic appropriate realization of the empty
   matrix concept is effectively utilized to transparently handle
   static and single vector input, single vector output systems within
   the more general context of dynamic, two vector input, two vector
   output systems.  The deficiencies of the present MATLAB realization
   of this concept are described. ]  )

_____________________________________________________________________
Andy Adler,    | Pulmonary Physiology Unit         | Lab 303-398-1626
adlera at njc dot org| | National Jewish Center,Denver,USA  | National Jewish Center,Denver,USA | Fax 303-398-1607

Just because you're not paranoid doesn't mean they're not out to get you