Re: [boost] Preview 2 of a template Geometry Library

From: Joel de Guzman (joel_at_[hidden])
Date: 2008-03-26 18:16:44

Phil Endecott wrote:
> Barend Gehrels wrote:
>> About the "point concept". I've looked at the concepts, downloaded and 
>> tried the concept compiler.
> 
> 
> Hi Barend,
> 
> Other people have expressed a desire that you describe your library in 
> terms of concepts I agree with that, see e.g. John Femiani's messages.
> 
> Based on the comment quoted above I wonder if perhaps you have 
> misunderstood what is being asked for.  I don't think anyone is asking 
> for code that will run on ConceptGCC.  I would be quite happy for you 
> to simply write your _documentation_ in terms of concepts.

Exactly!

> If we take your point class as an archetype of your point concept, I 
> think that this means that algorithm implementations have to use the 
> operator[] notation while "user" code can choose to use .x()/.y() 
> notation.  For what it's worth, as a potential algorithm author I'm not 
> at all enthusiastic about that: 

Me too.

> but I know that [] is other peoples' 
> preferred style.  

[integer] implies that the elements have the same type.

> No doubt once we look at the concepts for everything 
> else there will be myriad other similar choices.  

Yes. Fusion style at_c<N>(p) comes to mind.

> I fear what it comes 
> down to is this: if you present a minimalistic library, people will 
> focus on and disagree about these details; neither side of the ".x vs 
> [0]" argument is "right", and I don't think there's a solution that 
> keeps both happy.  

Yes.

> Only by presenting a library that has compelling 
> merit of its own (for example within its algorithms) will you get 
> people to set aside their stylistic preferences and follow your concepts.

Adaptability with existing and future point classes and geometry/
graphics libraries is a compelling point; and one that should
not be taken lightly.

Regards,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net