[boost] Re: Stdint header?

From: Kevlin Henney (Kevlin.Henney_at_[hidden])
Date: 1999-07-28 15:06:32

The alternative is to provide a template based solution, rather than death
by a thousand typedefs/#defines. I did something like this, and outlined
the details in a couple of articles for Overload when Sean was editing it a
few years ago. I'd be happy to dig it up and submit the code as a library.

Thoughts?

Kevlin

Beman Dawes <beman_at_[hidden]> on 28/07/99 17:38:46

Please respond to boost_at_[hidden]

To:   "Boost.org mailing list" <boost_at_[hidden]>
cc:    (bcc: Kevlin Henney/QA Training Ltd)
Subject:  [boost] Stdint header?

Background:  The new C9X C Langauge standard has added a header
<stdint.h> which declares integer types with specified widths and
corresponding MIN/MAX macros. The typedefs have names like int32_t,
int_least32_t, int_fast32_t, and intmax_t.  The C++ standard doesn't
supply this header.

The Question: How can Boost authors access specified width integers?
(This isn't theoretical; there are at least two boost projects
underway that need specific integer widths.)

Possible answers:

1)  #include <stdint.h>, and supply a version for systems which don't
already have one.  Cons: The names go in the global namespace, and it
seems like a C rather than C++ solution.

2)  #include <cstdint>, and supply a version (with names in namespace
std) for systems which don't already have one.  At least one compiler
vendor is already doing this.  Pros: Minimum future impact if C++
eventually standardizes <cstdint>. Cons: The C++ committee may never
standardize <cstdint> or may standardize it in a way incompatible
with C.  (These seem low probability, or even very low probability,
events, IMO.)

3)  #include <boost/stdint.hpp>, and supply same.  Names in namespace
boost.  Cons: If <cstdint> becomes either common or standardized,
Boost is out-of-step.

4)  #include <boost/our_name.hpp>, and supply same.  Namespace boost,
and use names different from <stdint.h> names.  Cons: Why fight a
standard? Boost should support existing standards, even if a bit
uncomfortable.

5) and 6)  Too brain-dead to bother explaining.

At first, I only considered (3), but then in came a compiler release
with <cstdint> supplied.  It took some getting used to, but now (2)
is starting to look better.  I don't really like (1) or (4) for the
reasons stated.

Comments?

--Beman

------------------------------------------------------------------------
Click Here to apply for a NextCard Internet Visa and start earning
FREE travel in HALF the time with the NextCard Rew_at_rds Program.
http://clickhere.egroups.com/click/449

eGroups.com home: http://www.egroups.com/group/boost
http://www.egroups.com - Simplifying group communications