[boost] Re: Thoughts on Roman numbers

From: Gennaro Prota (gennaro_prota_at_[hidden])
Date: 2002-11-01 07:37:04

On Fri, 1 Nov 2002 01:00:25 -0500, Daryle Walker <dwalker07_at_[hidden]>
wrote:

>On Saturday, October 26, 2002, at 1:51 PM, Beman Dawes wrote:
>
>> At 02:42 AM 10/25/02, Daryle Walker wrote:
>>
>>> As another poster on this subject has said, Roman numbers don't have
>>> different math rules, they are just a different external 
>>> representation
>>> of natural numbers.  This suggests that the place for new code is in
>>> the I/O system!  We'll use a special locale with custom facets.
>>
>> Boost libraries shouldn't always be dead serious. There should be a 
>> place
>> for fun libraries too.  They can be useful, too, and often make good
>> teaching examples.
>
>I didn't imply anything about the seriousness or levity on this 
>subject.  But having a "fun" project doesn't mean you should be lax on 
>where to approach the problem.

Agreed. But since the roman representation is likely to be used
sporadically or at least in a few well-defined places of an
application, mostly intermixed with the normal representation, I would
not go through the create locale/imbue it onto the stream/re-imbue the
old one every time I want a little number somewhere. I would find it
quite overkilling (and awkward) even with the I/O state savers. My
preference is for a little class to be used like:

 cout << roman_rep (200) << ...; // constructs a temporary and output

Simple and effective IMHO. Note that, as I said, once you make a few
design decisions the core of the code is just a 10 minutes exercise.

Genny.