[boost] [optional] Update II

From: Fernando Cacciola (fernando_cacciola_at_[hidden])
Date: 2003-10-27 19:18:54

Hi Joel,
Hi Roland,
H All!

Thanks to reports from Roland, Joel and Peder, I think I have solved the
regressions with the VC compilers.
In this update I'm also addressing some issues raised before:

1) References are now assignable (this was requested by Joel)

2) New headers: "detail/none_t.hpp" and "utility/none.hpp" were added
providing a new type 'detail::none_t' and a new global constant
'boost::none' which allows you to use a more value-like interface,
supporting the following:

optional<T> opt ;
if ( opt != none )
{
  // opt is initialized
  foo(opt);
}
opt = none ; // opt is reset() now.

NOTE: Boost.Optional itself doesn't care what is 'none_t' exactly as long as
it is a (distinct) type. So for the moment I wrote it as trivially as
possible. Discussions about the concept and definition of 'none_t' itself
can be done separately and in the future when concrete uses pop up.
The headers are split to allow Borland uses to precompile optional.hpp in a
.csm without the compiler complaining about initialized data (the global
instance 'none' is in a separate header that must be included explicitely)

3) OptionalPointee comparisons are now factored out in
"utility/compare_pointees.hpp" so that the concepts can be used
independently of optional<> itself (i.e., without having to include
"optional.hpp")

I've tested this with VC7.1
Unfortunately, since I have quit my old job (I'm freelance now) I'm left
with VS .NET 2003 only (by now), so I can't test it with any other
compilers.
I'll be looking at the compiler status tables.

I need to update the documentation, though, which I will do as soon as I
finish my second freelance project in about a couple of weeks.

Best regards,

Fernando Cacciola
SciSoft