CGAL users discussion list

[Efraim Fogel <>]

Applying callbacks or use visitors or observers does not come for free.
It depends on the necessity and the overhead. The decision whether to
have them is made separately for each alg. For example, the interface
for the sweep-line computation and zone computation (in the
Arrangement_2 package) allows the user to provide her/his own visitor
essentially implementing any alg. that is based on one of these 2
frameworks. The package provides several visitors, e.g., a visitor to
insert a collection of curves in an Arrangement_2 structure, a visitor
to compute the intersections of a collection of curves, etc. In
addition, one can attach observers to an instance of an Arrangement_2
structure, which results with a callback for each change of the
structure (e.g., the creation of a new face).

The operations provided by the Boolean_set_operations package are based
on the Arrangement_2 structure and can be applied using an underlying
arrangement that have observers attached to it.


 wrote:
> Hello everybody,
>
> A a user of CGAL, I wish to bring up a point(_3) which I think is important 
> to users of geometric software to which they don't want or need to change 
> to code itself. 
>
> I find it bothersome that I can't have more control over the process of the 
> geometric algorithms. For instance, if I want to compute the straight 
> skeleton, and in which to apply some function in every event (like matching 
> between the vertices participating or something), I cannot do this online, 
> and have to settle for post-processing, or some sophisticated workaround 
> which might be available, but I don't know about..
>
> Is there a way (or a possibility) to establish a way in CGAL to allow users 
> to pass functions as arguments to the algorithm, which it will apply in 
> specified occasions? (another example, the finding of an intersection in a 
> symmetric difference, so I wouldn't have as an output the same vertex in 
> two different polygon, and have to post-establish that this is the same 
> vertex).
>
> I think this functionality is very important, and I would like to hear you 
> opinions about it.
>
> Thanks,
> Amir Vaxman (CGGC Technion)
>   
-- 
  ____  _        ____             _
 /_____/_) o    /__________  __  //
(____ (   (    (    (_/ (_/-(-'_(/
                        _/