Re: [Boost-users] [Fusion] Adapted struct with member optional<variant<A, B>> needs typedef

Subject: Re: [Boost-users] [Fusion] Adapted struct with member optional<variant<A, B>> needs typedef
From: Christopher Schmidt (mr.chr.schmidt_at_[hidden])
Date: 2010-09-27 14:26:26

PB schrieb:
> On Mon, Sep 27, 2010 at 5:19 PM, Nathan Crookston
> <nathan.crookston_at_[hidden]> wrote:
>> Hi Pete,
>>
>> On Mon, Sep 27, 2010 at 5:28 AM, PB <newbarker_at_[hidden]> wrote:
>> <snip code>
>>> BOOST_FUSION_ADAPT_STRUCT(
>>>    Transition,
>>>        (boost::optional<boost::variant<Word,Text>>, token)
>>> )
>> <snip code>
>>> Compiler output:
>>> error warning C4002: too many actual parameters for macro
>>> 'BOOST_FUSION_ADAPT_STRUCT_FILLER_0'.
>>> <etc>
>>
>>> Why is this, and what are the constraints on the types fed to
>>> BOOST_FUSION_ADAPT_STRUCT?
>>
>> The problem seems to be that the preprocessor is seeing the ',' in
>> boost::variant and is parsing it like a tuple of 3 objects:
>> "boost::optional<boost::variant<Word," "Text>>" and "token", while
>> Boost.Fusion expects 2 objects.  The normal solution is to add a set
>> of parentheses around the token with a comma, but that doesn't seem to
>> work here.  I think your typedef solution is probably the simplest and
>> best for this case -- but my skills with the preprocessor are
>> definitely incomplete.
> 
> Nate,
> 
> Ahh. I didn't think about the comma. Perhaps the
> BOOST_FUSION_ADAPT_STRUCT macro itself needs some parentheses
> internally.
> 
> I'll stick to the typedef but hopefully if this can be fixed, the
> right people are viewing this topic.

I think this issue cannot be fixed. Is there a method to extract a type
from a parenthesized type expression without losing cv-qualifiers?
BTW. this issue already came up a few times. Paul Mensonides summed it
up pretty well 7 years ago:

http://article.gmane.org/gmane.comp.lib.boost.devel/84424

-Christopher