Re: [Boost-users] [fusion] probably a bug in result_of::join

Subject: Re: [Boost-users] [fusion] probably a bug in result_of::join
From: Christopher Schmidt (mr.chr.schmidt_at_[hidden])
Date: 2009-10-14 11:09:26

comments inline...

jl schrieb:
> Christopher Schmidt writes:
>> Yeah, this is a major issue of the current fusion implementation. The 
>> result_of::-metafunctions do not regard that the actual functions take 
>> const-qualified arguments only. On top of that the documentation is 
>> outdated and misleading.
> 
> Well you could make functions, metafunctions and doc agree like this:
> 
>    namespace result_of
>    {
>        template<typename LhSequence, typename RhSequence>
>        struct join
>        {
>            typedef joint_view<
>              typename add_const<LhSequence>::type,
>              typename add_const<RhSequence>::type
>            > type;
>        };
>    }

That's what probably best. This change would not break backwards 
compatibility and it would add a fair amount of comfort to fusion. I am 
in favour of adding this to the trunk.

> 
>> My C++0x port of fusion introduces non-const arguments for the 
>> algorithm functions and a new result_of::-concept which allows passing 
>> (optionally) cv and ref qualified types.
> 
> Do you plan to support that in the version for the "old" compilers as well?
> J-L

Of course. In fact, the code base is pretty much the same for both, 
c++0x and non-c++0x supporting compiler.
At the moment, compilers such as the gcc 3.x line and MSVC >=7.1 are 
already fully supported.

-Christopher