Re: [boost] Re: Dimensional Analysis Interest?

From: Jeff Garland (jeff_at_[hidden])
Date: 2003-10-10 09:24:15

On Fri, 10 Oct 2003 09:51:53 -0400, Jeff Flinn wrote

> Would you see the Date-Time library being subsumed by a "Units of Measure"
> library? 

No, not at all.  I've never seen a 'units' library that would deliver all the
specialized features that date_time does.  For example, things like
month_iterator, year_iterator, time_periods, clocks to measure time, time zone
adjustments, etc don't fit into what most people think of as units.  In fact,
most of what I have seen in the current units libraries boils down to the
time_duration class from date_time.

That is not to say that there shouldn't be inter-operation between these
libraries and perhaps some common core.  As an example, in date_time there is
an int_adpater template which allows integer types to support 'special values'
such as infinity, negative infinity, and 'not a number'.  I could see wanting
to have a small units type that supports this same sort of behavior. 

> I'm sure a lot was learned in that endeavor. What difficulties
> would you foresee? My intuition is that the complexities associated with
> date-time measures are on a similar order for the other fundamental
> quantities of mass, length,...

Perhaps.  I think that SIUnits, for example, benefits from an international
standard and well defined rules.  Many concepts in the date_time domain, such
as leap seconds, are defined by legislation instead of physics and hence
create lots of difficulties.  You can read more about this here:
http://www.crystalclearsoftware.com/libraries/gdtl/user_docs/Tradeoffs.html

Jeff