Re: [boost] [gsoc] Proposal - Boost.Matrix - DRAFT RFC

Subject: Re: [boost] [gsoc] Proposal - Boost.Matrix - DRAFT RFC
From: Emil Dotchevski (emildotchevski_at_[hidden])
Date: 2009-04-01 03:22:04

On Tue, Mar 31, 2009 at 10:25 PM, Patrick Mihelich
<patrick.mihelich_at_[hidden]> wrote:
>> IMO the abstraction provided by a template <typename T,size_t
>> Rows,size_t Cols> class matrix is unnecessary and will only get in the
>> way for computer graphics applications. I've seen several attempts
>> like this at generalizing matrices for game engine use getting
>> scrapped and refactored to a more traditional set of types (not
>> templates) and functions (this doesn't mean that such generalizations
>> are not appropriate in general.)
>
> And maybe it's not surprising that these attempts failed. It's not that easy
> to get the generalization exactly right. In a production environment it
> makes sense to just write some ad-hoc types rather than spend months on the
> perfect matrix class;

There is no such thing as the perfect matrix class. Also, there is no
such thing as the perfect fixed size matrix class. Matrices are great
generalization in mathematics; for programming, you need to start with
the intended application.

> Anyway I personally could care less about computer graphics, fixed-size
> matrices are also useful in computer vision, robotics, and no doubt other
> domains.

Maybe so, I don't have that much experience in those fields. What
prompted my response was that realtime computer graphics was stated as
one of the intended applications. In my opinion a different, less
abstract interface would be more appropriate there.

Emil Dotchevski
Reverge Studios, Inc.
http://www.revergestudios.com/reblog/index.php?n=ReCode