Empty matrices

[John W. Eaton]

On 15-Oct-2008, LUK ShunTim wrote:

| Hi,
| 
| This snippet is taken from some matlab code:
| 
| fd=inline('sqrt(sum(p.^2,2))-1','p');
| 
| p=[0.01 0.00; -0.51 -0.05; 0.01 -0.20] % NG!
| %p=[1.01 0.00; -0.51 -0.05; 0.01 -0.20] % Works
| %p=[1.01 0.00; -1.51 -0.05; 0.01 -1.20] % Works
| 
| d=feval(fd, p)
| ix=d>0
| a=d(ix)
| 
| val=0.05;
| b=feval(fd, [p(ix,1)+val, p(ix,2)])
| 
| res=b-a
| 
| It works when ix returns at less one non-zero index but fails if ix=0.
| The result of using p=[0.01 0.00; -0.51 -0.05; 0.01 -0.20] is:
| 
| ix =
| 
|    0
|    0
|    0
| 
| a = [](0x0)
| b = [](0x1)
| error: operator -: nonconformant arguments (op1 is 0x1, op2 is 0x0)
| error: evaluating binary operator `-' near line 14, column 6
| 
| I cannot claim that the matlab code works as I've no matlab installed to
| do the test, though. :-(

I think the dimensions must always conform unless one of the operands
is a scalar, in which case the operation with the scalar is applied
to all elements of the non-scalar operand.

| How to proceed?

Fix your code.

| Or is this a bug in octave?

I don't think so.

jwe