Re: Explanation of Octave operator overloading

[John Swensen]

David Bateman wrote:
> John Swensen wrote:
>> Could someone give a quick tutorial on how the operator overloads work
>> in Octave (and maybe add it as a section in the Da Coda Al Fine section
>> of the wiki)?
>>
>> For example, I have defined a new type that derives from
>> octave_base_value called octave_socket.  The problem is that I want to
>> be able to compare it to numerical values.  I have already tried simply
>> overloading the operator< to compare against octave_value.int_value() as
>> well as just a plain int.
>> I see there are a bunch of macros called OCTAVE_INSTALL_*_*_*_CMP_OPS
>> and assume I have to do something with these, but haven't been able to
>> figure out what.
>>
>> Thanks.
>>
>> John Swensen
>>
>>   
> A good little example is in
> octave-forge/extra/linear-algebra/ov-re-tri.cc which overloads the
> assign, ldiv, hermitian and transpose operators for a triangular
> matrix type. The reason its good is that it is tiny. In particular
> check the install_tri_ops functions, the functions that are defined
> immediately above this for the overloaded functions, and finally note
> that install_tri_ops is called immediately after the type is
> registered in the replacement chol.cc function. Note that I intend to
> delete this code in octave-forge now that the code to treat positive
> definite matrices is in octave itself, so rescuing it and using it as
> an example in da coda or del segna is probably a good idea. In fact if
> you are really motivated, rescuing both da coda and del segna, and
> this code and writing a chapter on oct-files for the octave 3.0 manual
> might be even better :-)
>
> Regards
> David
>
>
Thanks for pointing me in the right direction.  That example gave me
exactly what I needed.  I did find, however, that for simple comparison
operations, there are macros to make it *really* easy.  The DEFBINOP_OP
macro allows you to simply implements an accesor function in the
user-defined type to return some integer value that can then be compared
against other octave types.  In my case, I have a octave_socket class.
I used the DEFBINOP_OP macro as follow:
DEFBINOP_OP (lt, scalar, scalar, <)
DEFBINOP_OP (le, scalar, scalar, <=)
DEFBINOP_OP (eq, scalar, scalar, ==)
DEFBINOP_OP (ge, scalar, scalar, >=)
DEFBINOP_OP (gt, scalar, scalar, >)
DEFBINOP_OP (ne, scalar, scalar, !=)

I then had to overload the scalar_value() method in my user-defined type
and now I can compare against integers with all the operators.

Thanks again!

- John Swensen