Bug or feature?

[John W. Eaton]

On 20-Jan-1997, Shmuel Tomer <address@hidden> wrote:

: In matlab:
: 
:    >> [1 2] == [1]
:  
:    ans =
:  
:         1     0
:  
:    >> [1 2] == []
:  
:    ans =
:  
:         0
:      
: but in octave:
: 
:    octave:1> [1 2] == [1]
:    ans =
:  
:      1  0
:  
:    octave:2> [1 2] == []
:    error: operator ==: nonconformant arguments (op1 is 1x2, op2 is 0x0)
: 
: Is this a bug or a feature?

I think I would call this a bug.

: In other words, would it be correct to reverse the order of the
: following two tests in the macro MX_MX_BOOL_OP (src/ops.h, lines
: 94-100)? 
: 
:       if (m1_nr != m2_nr || m1_nc != m2_nc) \
:         { \
:           gripe_nonconformant ("operator " op, m1_nr, m1_nc, m2_nr, m2_nc); \
:           return Matrix (); \
:         } \
:       if (m1_nr == 0 || m1_nc == 0) \
:         return empty_result; \
: 
: Would that have (disastrous) effects, other than to provide a Matlab
: compatible behavior?

I don't think that simply switching the order of the tests is enough
to make this part of Octave compatible with Matlab.  I've just posted
a patch to the bug-octave mailing list that might do a better job.

: I don't prefer this behavior per se, but being able to use octave in
: my site may depend on wheter the users need to go through the
: existing zillion scripts and change every occurrence of 
:
:       if myString == '' 
: to
:       if strcmp(myString, '')

The patch will also make

  myString == ''

work as you expect.

Thanks,

jwe