Re: [boost] Interest in dimension/unit-checking library?

From: Deane Yang (deane_yang_at_[hidden])
Date: 2005-10-14 11:05:14

Andy Little wrote:
> "Matt Calabrese"  wrote
>> This is where I disagree. Radians ARE just ratios. More specifically,
> 
> [cut]
> 
> The definition of a radian in Appendix A (A.9) of the Guide for the Use of the
> International System of Units is:
> 
> 
> "The radian is the plane angle between two radii of a circle that cut off on the
> circumference an arc equal to the radius."
> 
> 
> You need to accept that definition if you are to have any chance of writing a
> credible Units library I'm afraid.
> 

I'm afraid I agree with Andy and others who have posted on this. I don't 
know of a single mathematician who views "radians" as a synonym for 
"ratios", and although wikipedia is not bad for many mathematical 
concepts, it is definitely not authoritative. But it's more than that. I 
don't know how to define a clear set of semantic rules (what operations 
are allowed or are not allowed) for pure ratios, and I have not been 
able to follow what Matt has in mind here.

You (Matt) might have a case, but you haven't convinced us yet. I would 
suggest that for the moment you represent ratios of the same quantity by 
a pure number type (I hope this will be a template parameter, because I 
want to use something besides float or double), but you should feel free 
to keep making your case. A short summary of what the allowable 
operations are, analogous to

point + vector = point
point - point = vector
vector * scalar = vector
vector +/- vector = vector

would be helpful. To be honest, the "rules" that you've tried to come up 
with so far seem ad hoc to me and therefore unlikely to work in general. 
Here's an example: it is useful to feed decibels into the "10^x" 
function and radians into the sine function, but it is not useful to 
feed decibels into the sine function or radians into the "10^x" 
function. How do you propose to handle that kind of situation?