From: Darin Adler (darin_at_[hidden])
Date: 1999-11-28 19:59:39
> This has one disadvantage that I know of:
> 1) It requires a working version of <limits>.
One more big disadvantage:
2) It will pass a UDT by value if that UDT is a numeric type and
numeric_limits is specialized for the type.
I think that's a fatal flaw in my proposed change.
This makes me think of the "type traits" proposal. I do want to avoid the
pitfall of a single catch-all traits template, but I'd also like to avoid
listing "bool, signed char, unsigned char, short, unsigned short, wchar_t,
int, unsigned int, long, unsigned long" more than once. Once reason is that
I want to add "long long" to that list as compilers add support for it (due
Things like is_scalar, is_array, is_pointer, is_reference may also be useful
and could perhaps go into something like type_traits without creating a
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk