From: Jan Langer (jan_at_[hidden])
Date: 2003-10-24 18:17:03
Hugo Duncan wrote:
> if we have
> quantity<NumericType, Unit, Dimension>
> then we can have a trait to determine whether or not the Unit has any
> sort of trnsform asociated with it. The Unit could then provide the
> (possibly static) functions required to perform the transform.
in general i think we should stick to the idea that everything is stored
and calculated in the fundamental unit. it should be made easy to change
the fundamental unit. then everyone who needs a different internal
scale, must say beforehand that he wants to use such a system. something
typedef units_system <si, light_years> my_system;
quantity <double, my_system::length> l1 (12); // 12 light_years
// something quite small
quantity <double, my_system::length> l2 = 1 * meter;
if the other path is followed, i'm also strongly in favor of hugo's
idea, because it allows to do more complicated conversions between
units. it could for example solve the celsius/fahrenheit issue.
but a problem might be to avoid unneccessary conversions. for example
between two non-fundamental conversions like km to inches. both units
know their conversion to meters. but i dont know if it might be an issue
if it multpiles with 1000 and then divides by 0.0254.
and what should be the return type of an addition of kilometers and inches.
-- jan langer ... jan_at_[hidden] "pi ist genau drei"
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk