[Scilab-Dev] problem with function overloading

[StŽéphane Mottelet]

Samuel Gougeon a écrit :
> Hi Stéphane,
>
> ----- Message d'origine -----
> De : StŽéphane Mottelet
> Date : 02/03/2010 09:35:
>> Hi all,
>>
>> I have some difficulties to understand how overloading of functions 
>> (not operators) works.
>> E.g. I would like to redefine delete() for handles, which is done 
>> actually with %h_delete.
>> (BTW, when I look at the source i see some calls to delete with 
>> handle arguments, how does
>> scilab cope with apparent recursivity here ?). When I try to redifine 
>> %h_delete in a
>> library loaded at startup time, then the original %h_delete is still 
>> used instead of the
>> "new" one. How can solve this problem (I also need to redefine %h_set 
>> and %h_get...)
>>
>> thanks for help
>>
>> S.
> AFAIU, overloading enables (and is aimed) to _extend_ a function in 
> order to
> handle _new data types_, not to redefine it to the already supported 
> data types.
> By default, whether arguments are of the "default" supported data 
> types, the
> regular definition of the primitive is used and applied. Otherwise, 
> and only then,
> the overloading definition is applied.
>
> There is no reason that SCI/modules/gui/%h_delete exists, since the 
> "h" = graphic
> handle data type is the default one for the delete() primitive. So, 
> %h_delete likely
> cannot be called by scilab, unless maybe it is aimed to be used with a 
> _vector_ of
> handles instead of a scalar handle (whether the primitive is defined 
> only for a scalar
> handle as input argument).
>
> For redefining a primitive, you should overwrite it with a macro.
> Now, if you want to use the initial definition of the primitive inside 
> its own redefinition,
> you should first define an alias of the primitive. Here is an example 
> for the max()
> function. Its default definition returns the position only of the 
> first occurence of the
> max value, while i need the position of all occurences. Therefore, i 
> want to redefine
> max() as well as to use the original max() primitive inside it. I will 
> do :
>
> m=grand(10,5,"uin",0,9) // Create data for test
> [v,ij]=max(m) // test before redefinition of max()
> Max=max; // Creates an alias.
> deff("[v,k]=max(m)","v=Max(m); k=ind2sub(size(m), find(m==v));") // 
> redefines max() using the alias
> [v,ij]=max(m) // Applying the new max()
> [v,ij]=Max(m) // Applying the former max()
> // max=Max; // Restoring the old max()
>
> Note that it is not possible to create the alias inside the 
> redefinition, since
> the definition is overwritten prior to be called ;)
> Then, the Max=max; may be done for instance in the scilab.ini (i did 
> not try).
>
> HTH
> Samuel
>
> PS: Because overloading is not aimed to redefine, i am not sure that a 
> bug report is relevant.
>
Hi Samuel,

I already know this trick; the problem is that it it does not work with 
overloading functions, that
is what I was explaining in the bug report.
I fully understand that overloading is not aimed to redefine, but this 
is the only solution for me
to emulate new properties for graphic objects (in the Plotlib). I use 
the user_data field.
You will sayt that I could have directly used a tlist, and overload set 
and get, but the big problem in Scilab is
that tlists are not passed by reference in macros, although graphic 
handles are...

So, I need to modify %h_set and %h_get. The problem is that they do not 
seem to be
overwritable; this is what the bug report is about.

S.