Re: [eigen] two decisions to take

[Benoît Jacob <jacob@xxxxxxxxxxxxxxx>]

On Tuesday 04 September 2007 21:33:27 Andre Krause wrote:
> > It's not a big deal adding member functions. In Eigen1 we allowed both
> > dot(v,w) and v.dot(w), so I think we'll do the same in Eigen2, since
> > there seems to be strong advocates of both approaches :)
>
> fine, as long as the global dot or cross functions are present (of
> course inside eigen's namespace), i have no objections against member
> functions. if it is well documented that those change the value of the
> object itself.

wait, I didn't mean that! u.cross(v) should return a cross-product expression, 
not assign it to u !

> too bad c++ cannot overload the dot-operator... but well, if "|" is that
> well known in quantum mechanics (to be honest, never seen that use, do
> you have some link / pdf showing that?), it sounds fine to overload it
> for dot product.

Actually it's not just quantum mechanics, it's really common in mainstream 
mathematics. The dot product of x with y is often denoted by (x|y) or <x|y>, 
though the most common form is <x,y> (but we don't want to overload operator, 
here).

See this page:
http://en.wikipedia.org/wiki/Bra-ket_notation
especially the bottom:
http://en.wikipedia.org/wiki/Bra-ket_notation#Notation_used_by_mathematicians

>
>
> kind regards, andre!
> p.s. i am very excited about the upcoming new eigen! thanks for all your
> efforts, benoit ! i like your way of "keep it as simple as possible". so
> maybe even c++ noobs like me will have a chance to understand the magic
> behind expression templates. or at least get the rough idea.

Thanks for the kind words. I didn't know either about ETs before Ben mentioned 
them, and I would not have understood them if I hadn't spent time playing 
with TVMET's source code. For Eigen2 I will write documentation about them, 
because I want to lower the barrier to contribution.

Cheers,
Benoit

Attachment: signature.asc
Description: This is a digitally signed message part.