Re: [boost] Formal Review: Boost.Polygon

Subject: Re: [boost] Formal Review: Boost.Polygon
From: Fernando Cacciola (fernando.cacciola_at_[hidden])
Date: 2009-09-15 12:00:55

Hi Luke,
>> 
>> ** The polygon_data type should be mutable (i.e. support things like 
>> push_back()), or a mutable version should be provided, or
>> std::vector/list<point_data> should work with minimal traits
>> specialisation.
> 
> The example custom_polygon.cpp shows the traits for a list<CPoint>.  I could
> make all containers of points models of polygons_concept by default, but I
> wouldn't know which polygon concept the user prefers.
> 
A mere container of points should be a model of the most general polygon concept 
(which is polygon_concept)

>> I also found that polygon_data is specialised by the coordinate type, not
>> the point type, so anyone using their own point type will have to provide
>> their own polygon type. Is there some reason why I can't have
>> polygon_data<my_point>?  Again, making it easier to use
>> std::container<my_point> would help.
> 
> Point type is a trait.
> 
Can you elaborate in your response?

If I have

struct my_point { float x,y; }

How do I create a "polygon_data" that uses that point type?

AFAICT that is just not possible.

> 
>> ** extents() and center() should return their results, rather than using an
>> output reference.
> 
> Out parameters and equational reasoning have been discussed elsewhere in the
> review.
> 
Yes, but I'm not sure what is your position regarding that.
Can you clarify it?

Best

--
Fernando Cacciola
SciSoft Consulting, Founder
http://www.scisoft-consulting.com