Re: [Boost-users] [Any] inconsistent return types of any_cast overloads

Subject: Re: [Boost-users] [Any] inconsistent return types of any_cast overloads
From: Michael Chisholm (chisholm_at_[hidden])
Date: 2013-12-04 14:16:16

On 12/4/2013 7:07 AM, Krzysztof Czainski wrote:
> 2013/12/4 Arne Vogel <avogel_at_[hidden]>
>
>>   On 03.12.2013 00:01, Krzysztof Czainski wrote:
>> [...] what do you gain by doing:
>>
>>    const auto& vec = boost::any_cast<std::vector<int>>(makeVec());
>>   instead of
>>    const auto vec = boost::any_cast<std::vector<int>>(makeVec());
>>   ?
>>
>>   Regards,
>> Kris
>>
>>
>> In C++11 with a moveable type, probably not much. I don't quite remember
>> why I wrote the code like this - it could have been out of habit since in
>> C++98 using a reference may save a copy, though I guess an optimizing
>> compiler had already been allowed to elide it.
>>
>
> I'm just curious to know if anyone uses a compiler that doesn't actually
> elide the copy (or move) in this case ;-) In all discussions about this
> I've never seen someone actually name one, but still they try to avoid the
> copy just in case.
>
> I think taking values returned by value by const & to extend their lifetime
> should be deprecated in new C++ - compilers could then warn not to do it.
>
> Cheers,
> Kris
>

An old article comes to mind which uses exactly this idea to implement a 
little API to help people write exception-safe code [1].  This is part 
of the Loki library [2].  I think this is still considered a legitimate 
technique, right?

Andy

1. 
http://www.drdobbs.com/cpp/generic-change-the-way-you-write-excepti/184403758
2. http://loki-lib.sourceforge.net/