Re: [boost] Formal Review for Interval Library

From: Guillaume.Melquiond (gmelquio_at_[hidden])
Date: 2002-09-10 17:21:11

On Tue, 10 Sep 2002, Fernando Cacciola wrote:

> Great!!
> Where can I get the update code from?

Sorry, it isn't available for the moment. As I explained, we don't use
Windows computers, so it is always a bit painful to test new
modifications. And since it's already very late here, I won't do it now.

However, if you really want to test it, just comment out the problematic
lines :-). Indeed, they are not really important if you just want to have
a shot at the library. The problems are the bodies of the constructor
(which test for exceptional cases; I really don't understand why the
compiler isn't happy); and BCC seems to lack a 'rint' function.

> Hmmm.
> I'm confused.
> Is there any reason why the policy don't have to handle sqrt_down/up(0)?
> What does it mean that it 'don't have to handle'? Should I expect:
> sqrt_up(0)==-3.14? Or that it throws?

Not at all. It's just that 'sqrt_something' will never be called with an
argument equal to 0. It will only be called with a strictly positive
argument. So, when the argument is 0, even if you return -3.14 or throw an
exception or destroy the universe, it won't change the behavior of
'sqrt(interval)'.

Regards,

Guillaume