Re: [Boost-users] [units][geometry] subtract_typeof_helper<> help!

Subject: Re: [Boost-users] [units][geometry] subtract_typeof_helper<> help!
From: alfC (alfredo.correa_at_[hidden])
Date: 2010-12-31 21:35:13

Sorry to revive this old thread,

 First I would like to know if there had been any progress in this front. 

 Second, I would like to pick the attention of Boost.Geometry again in order 
to suggest them use boost.units as a way to make the Boost.Geometry library 
very general. 

 Third, after some thinking I have some thoughts about the original question 
in this post

What I'm struggling with is how to provide a units-aware Vector/Point system 
>
> for physical computation. 
> Currently, I'm favoring using units::quantity<U, Vector<double>> rather 
> than 
> Vector<units::quantity<U>>. 
>

I think none of the two options makes general sense. The only point at which 
units make sense is at the "metric" of the space. Since in Boost.geometry 
library coordinate system an metric are tied together, we need a special 
type of coordinates system. For example one suggestion could be to extend 
the coordinate systems of Boost.Geometry to include the units and dimension 
of space. For example one way to specify points in space could be by means 
of a type like this

 model::point<double, 3, cs::unit::cartesian<si::length> >

and then specialize all the template functions that use cartesian as 
coordinate system, for example distance. This makes sense because the only 
point at which units are useful is when we take distances, modulus or 
projections between vectors in that "space" with unit/dimension.

 

> I expected to use boost::geometry, but that didn't seem to provide much 
> benefit. 
>

Regardless of whether we see benefit or not at this point I think this is a 
real problem that we should face. This is a conceptual problem that must be 
solved sooner or later.
 

> Now that I'm not using boost::geometry to implement Vector/Point, it might 
> be much easier to support Vector<units::quantity<U>> as well. 
>

> My example was an intermediate stage.  The example file that I failed to 
> post looked like this, along with the sample output. 
> It is interesting that (somehow) the Point/Vector quantities figured 
> themselves out. 
> Its interesting because: 
>    Vector = Point - Point 
>    Point = Point + Vector 
>   Vector = Vector + Vector 
> but _not_ Point = Point + Point 
>   and the dot product 
>   Scalar = Vector * Vector 
>
> #include "BasicVector.h" 
>

can you please post what is your current implementation of BasicVector.h?

  cout << disp1 << endl; 
>   cout << abs(disp1) << endl; 
>   cout << abs_squared(disp1) << endl; 
>   p1 += disp1; 
>   cout << p1 << endl; 
>   Time dt = Time(0.1 * s); 
>   Velocity v = (p3 - p2) / dt; 
>   cout << v << endl; 
>   return EXIT_SUCCESS; 
> } // main 
>
> Output: 
> <3 3 3> m 
> (4 5 6) m 
> <-10 40 -20> m s^-1

this is the sort of things that from a dimensional point of view don't make 
sense. In the previous "m" or "m s^-1" is just a tag for the vector but it 
is not a real unit of the vector.
This is more ovbius when vectors are represented in spherical coordinates.

Any thoughts?

Thank you,
Alfredo