|
Boost : |
From: David Abrahams (abrahams_at_[hidden])
Date: 2001-01-20 14:00:57
----- Original Message -----
From: "Howard Hinnant" <hinnant_at_[hidden]>
> David Abrahams wrote on 1/19/2001 5:58 PM
> >It seems like a number of the templates aren't able to give good results
for
> >user-defined types. For example, shouldn't is_integral<T> check
> >numeric_limits<T>::is_integer when possible?
>
> Speaking only for myself, I view the type_traits structs as answering:
> What is this? But I see std::numeric_limits answering: How does this
> behave? So, for example given the class BigInt, I would expect
> is_integral<BigInt> to answer false, is_class<BigInt> to answer true, and
> numeric_limits<BigInt>::is_integer to answer true (assuming BigInt's
> author had specialized numeric_limits).
OK. /If/ that is the intent, I think we should be clear about it in the
documentation.
Also, as an aside, I think we need a (partial) specialization of
numeric_limits for boost::rational<T>.
-Dave
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk