| Re: [eigen] Eigen appears to rock. |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
Quoting Thomas Vaughan <tevaughan@xxxxxxxxx>:
On Fri, Aug 22, 2008 at 8:07 AM, <jacob@xxxxxxxxxxxxxxx> wrote:
Hm regardless of intuitiveness discussions... it remains that
p1+0.5*(p2-p1)
is slower than (p1+p2)*0.5 so it would be very unfortunate to require
our users to do that! Same thing with any approach based on converting
back and forth between vector and point.
Agreed.
The right solution is for me to base my Point class on your framework
because it works for me. :^)
I was about to mention that -- absolutely nothing prevents you from
doing that.
It seems too bad that there is no way to express in an efficient way
certain mathematical ideas, such as that two points in space have a
difference but no sum (as can be seen by the consideration of vectors
and points independent from their representation in a particular
coordinate system).
I know what you mean. Your idea of a Point class makes perfect
mathematical sense and is indeed better mathematical practice. I can
believe that it will even prevent bugs in your programs. But the
things work is unfortunately that best mathematical rigor is not
always compatible with best performance, binary code size, compilation
times, or API. So a compromise has to be made. Eigen's philosophy
means that the good compromise for it is to have only vectors and no
separate points.
Thanks for your patience!
No problem, thanks for writing.
Cheers,
Benoit
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.