[boost] Re: Unit library

From: Deane Yang (deane_yang_at_[hidden])
Date: 2004-01-11 16:04:40

Phil Richards wrote:

> On Sun, 11 Jan 2004 13:56:16 -0500, Deane Yang wrote:
> 
> My reply is those who have posted saying "You don't need to all this
> complicated dimensional analysis stuff - just build a units library then
> add dimensional analysis on top".  I don't believe it can be done that way
> round - which I think is what you are saying too.  I believe (and have
> kept saying) that a units library must incorporate dimensional analysis in
> order to be useful.  Others didn't appear (to me) to have that view. 
> Maybe I was confused.

To be honest, I think you were. As far as I can tell, everyone has
always assumed that the units library would handle the natural
derivation of units arising from arithmetic operations.
In particular, the units library should contain

template <class unit1, class unit2>
product<unit1, unit2>::type
operator*(unit1 quantity1, unit2 quantity2);

template <class unit1, class unit2>
quotient<unit1, unit2>::type
operator/(unit1 quantity1, unit2 quantity2);

template <class unit1, class rational>
power<unit, rational>::type
pow(unit base, rational exponent);

I think everyone agrees with this. The question is whether the
types "unit1", "unit2", and "unit" should be replaced by
"dimension1", "dimension2", and "dimension".

(By now, I think they should be dimensions with respect to explicitly
stated units, which is, in fact, what I think some of the proposals
actually do. You want to allow the same dimension but with different 
units, AND you want to allow different dimensions that use the same
units. A given type should specify BOTH the dimension and the unit.)