Boost logo

Boost :

From: Guillaume Melquiond (gmelquio_at_[hidden])
Date: 2002-09-06 05:03:48

On Thu, 5 Sep 2002, David Bergman wrote:

> Guillaume,
> What is the rationale behind *not* defining
> [a, b] == [c, d] as a == c && b == d
> And
> A <= B A < B || A == B
> ?
> This was suggested by both me and Joel.

Because such a definition has no practical usefulness (for what I
know) :-).

> I actually think the interval arithmetic should be separated, into
> 1. interval core definitions, including '<', '==' and set operations
> 2. arithmetical operations on intervals
> 3. non-arithmetical operations on intervals

Seems good to me. Gennadiy has already made such remarks.

I would even add a fourth part for the floating-point specializations;
since they are useless for somebody using interval< rational<... > > or
interval<mpfr> or interval<date>, etc.

> And there should be openings for distribution descriptors for
> probabilistic applications, as suggested by someone. That could be added
> as a separate extension (header file)

I'm completely dumb when it comes to probabilistic applications, so don't
expect me to do that :-). (I'm only speaking for myself and not for the
two other authors).

> This way, one could use (1) for those strange applications, such as date
> intervals, where no (advanced) arithmetics is needed ;-)
> Regards,
> David



Boost list run by bdawes at, gregod at, cpdaniel at, john at