RE: [boost] boost::rational test program

From: Moore, Paul (paul.moore_at_[hidden])
Date: 2001-01-22 05:14:22

From: Stephen Silver [mailto:boost_at_[hidden]]
> Paul Moore wrote:
> 
> > > http://www.argentum.freeserve.co.uk/cpp/rational.zip
> >
> > This is nice. Would you have any objections to me including it in
> > the rational library, as a regression test?
> 
> No objections.  I already gave it a Boost-compliant copyright notice.

Thanks. I noticed the copyright was OK, but I thought I'd better ask anyway.
I've made a few cosmetic changes as well (don't fail on the first error, and
report number of failures). I've also added a trivial user-defined integer
type (just wrapping int) which I can use to test the user-defined type
issues.

> > Your example
> >
> >     std::istringstream iss("57A");
> >     iss >> r0;
> >
> > consumes "57". This is a valid representation of a rational, namely 
> > 57/1, so r0 is set to this, and the stream is in a good state.
> 
> OK.  But in that case the documentation needs changing, because it
> says:
> 
>     The external representation of a rational is two integers,
>     separated by a slash (/). On input, there must be no space
>     between the numerator and the slash. 
> 
> which seems to imply that anything without a / is not a valid input.
> (Also, the above requirement seems lacking in symmetry.  Would it not
> be better if it either allowed spaces both sides of the / or 
> disallowed them both sides?)

Agreed. I spent a while trying to work out what was going on, and I never
thought to check back on how I'd documented it all :-) I'll tidy up as you
suggest. (Actually, I'm now wondering whether the numerator-only
representation is worth the confusion it causes. Any comments based on
real-life needs?)

Paul.