Re: [boost] Improving the assignment operators of various Boosttypes

Subject: Re: [boost] Improving the assignment operators of various Boosttypes
From: Michael Marcin (mike.marcin_at_[hidden])
Date: 2008-09-10 14:23:53

Peter Dimov wrote:
> Michael Marcin:
> ...
>> For rvalues yes but for lvalues you need move emulation. Which can be 
>> implemented in the sandbox move library as follows.
>>
>> class foo
>> {
>> public:
>>    foo();
>>    foo( T t );
>>    foo( boost::move_from<foo> other )
>>    {
>>       swap( other.source );
>>    }
>>
>>    foo& operator=( foo rhs )
>>    {
>>       swap( rhs );
>>    }
> 
> I see what you mean now, thanks. This is, I believe, the Adobe approach 
> for move emulation:
> 
> http://stlab.adobe.com/group__move__related.html
> 
> It's based on the Adobe/Stepanov philosophy that all types are copyable, 
> so it doesn't support move-only types. This takes away one motivation 
> for having move. std::vector won't use adobe::move/boost::move, so this 
> takes away the other motivation. What's left? :-)

Adobe may believe that all types are Regular but I have types that can't 
be copied (at least not efficiently) and I disallow their copying to 
prevent clients of my code from shooting themselves in the performance 
foot. The move library in the boost sandbox by Daniel James, which is a 
continuation of the work by others including Adobe, relaxes these 
constraints and supports move only types.

I use it in some performance sensitive code or when I need/want stronger 
exception safety guarantees.

Thanks,

Michael Marcin