Re: [boost] Boost Units library preview

From: Kevin Lynch (krlynch_at_[hidden])
Date: 2006-08-21 20:18:46

Deane Yang wrote:
> Eric Lemings wrote:
>>> This brings up a good question.  Say you have two objects:
>>>
>>> meters m = 1;
>>> feet f = 2;
>>>
>>> What is the type (or unit) of the following expression?
>>>
>> m + f;
>>
> 
> I apologize for repeating myself, but I want this to fail to compile. I 
> do not want implicit conversion of units. I believe I am not completely 
> alone in this view.

No, you're not  :-)   I too want this to fail; having the ability to 
explicitly forbid on a calculation by calculation basis would not help 
... forget it once, and you're screwed.

I know it has been a while, and apparently not many people were swayed, 
but we did discuss this issue at length a few months back.  Implicit 
conversions imply choosing a "canonical" basis (or unit categories: 
length, time, etc...) and system of units internally (SI, CGS).  Many of 
the potential users of this library have neither a need nor an interest 
in doing calculation in SI (think astrophysics ... meters are not 
useful!), and many of us don't have a use for the fundamental unit 
categories of the SI either (length and time are the same thing, darn 
it!).  This also leaves out monetary units and what we have called "unit 
tags".

Those of us who do lots of long running calculations in these non-SI or 
non-standard unit systems seriously can't afford the accuracy damage 
that loads of implicit conversions carry along -- I work very hard to 
structure my numerical expressions to minimize roundoff error, and 
uncontrollable implicit conversions under the hood would be very bad! 
Additionally, forcing a canonical internal system makes it prohibitively 
difficult to use non-standard bases (as I would like to do :-)

Which is not to say that implicit conversions aren't useful to some, I 
just think that it's almost always indicative of a design error in the 
application.  But what do I know :-)  Perhaps this calls for types that 
implicitly convert, and types that don't, so that you never use the 
dangerous types if you don't want them.

Oh, and I also want to be able to define my own bases and units and tags 
and .... I want it all!

-- 
-------------------------------------------------------------------------------
Kevin Lynch				voice:	[phone]
Physics Department			Fax: [phone]
Boston University			office:	 PRB-361
590 Commonwealth Ave.			e-mail:	 krlynch_at_[hidden]
Boston, MA 02215 USA			http://budoe.bu.edu/~krlynch
-------------------------------------------------------------------------------