Re: [Boost-users] problem... maybe boost could help

From: Lukasz (lkudra_at_[hidden])
Date: 2007-06-27 11:01:40

Hello again,

I actually have another question. Although the solution now works, it's
still not perfect. The problem is the actual variant. Right now, i have a
separate header file which has a typedef for my variant, i.e.

typedef boost::variant<FWK::AGENT::Event, 
                        FWK::AGENT::EventTest,
                        FWK::AGENT::EventBehaviourStarted>                
event_container;

Multiple files that are involved in the messaging service use this typedef.
It works perfectly well. However, in the long run, for every custom event
that is derived from Event, a framework user would have to go play around in
this header to add the custom type, which is definatly something i would
like to avoid (for obvious reasons). Since i will have to use mpl::vectors
anyway to construct my variant (there could possibly be hundreds of events),
i was wondering if there would be a way of constructing the vector "along
the way", i.e. if every Event derived class could add to the vector, and
then i could in the end construct my variant based on that. I have barely
started reading up on mpl, so i'm not really sure how to go about this. I
was thinking of something like

struct list_helper
{
    typedef boost::mpl::vector< FWK::AGENT::Event > type;
}

This would be my starter type. But how can i push new elements into this
vector, especially since all the calls would be in separate files ? I also
would have to make sure that the call to boost::make_variant_over would
always happen after all the events are added. I looked a little over the
serialization library's BOOST_CLASS_EXPORT, which i figured is "kind" of
what i want to do, but in my case, i think it's simpler. However, i'm still
stuck for now.

Thanks for any help in advance,

Lukasz K.

Ovanes Markarian wrote:
> 
> 
> If you like, prepare me some simple example and I can correct it or help
> you at this point.
> 
> 
> With Kind Regards,
> 
> Ovanes Markarian
> 
> On Tue, June 26, 2007 14:56, Lukasz wrote:
>>
>> You summed it up very well with the first paragraph. I've been trying to
>> template this from the beginning, but it just wouldn't fit into the
>> design.
>> However, i think i'll be able to use variant within one of my maps to
>> store
>> the "real" type of the event which should work perfectly for my needs.
>> Thanks for all the input... it really helped to get some ideas together
>> and
>> variant will probably do the trick (but i still have to fiddle some more
>> with the idea).
>>
>> Thanks again.
>>
>> Lukasz K.
>>
>>
>> Ovanes Markarian wrote:
>>>
>>>
>>> My idea is based on fact that you (program developer or another
>>> developer)
>>> know what is going to
>>> be send. What you are trying to do is erasing the type to some base and
>>> that would not work.
>>>
>>> So you send event must handle derived type and put it correctly to
>>> variant. A possible solution
>>> would be:
>>>
>>> template<class EventType_>
>>> void sendEvent(EventType_ const& ev)
>>> {
>>>    m_eventContainer = ev; //<- now compiler knows exactly the type of ev
>>> m_derivedHelper->dispatchEvent(&m_eventContainer);
>>> }
>>>
>>>
>>> Best Regards,
>>> Ovanes
>>>
> [...]
> 
> 
> _______________________________________________
> Boost-users mailing list
> Boost-users_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/boost-users
> 
> 

-- 
View this message in context: http://www.nabble.com/problem...-maybe-boost-could-help-tf3959613.html#a11326230
Sent from the Boost - Users mailing list archive at Nabble.com.