Re: [boost] [string] Yet another Unicode string class

Subject: Re: [boost] [string] Yet another Unicode string class
From: Anders Dalvander (boost_at_[hidden])
Date: 2011-02-14 01:10:57

On 20:59, Dave Abrahams wrote:
> On Thu, Feb 10, 2011 at 3:38 PM, Anders Dalvander<boost_at_[hidden]>  wrote:
>> * It is (currently) immutable and shares data, and thus fast to copy.
>
> Is there a reason you think this is a better approach than you could
> have by using the small string optimization and simply doing a deep
> copy when the internal buffer is too small?
>

Thanks for your comments.

Actually, the only reason I have for this approach is that I listened, 
perhaps too much, to other voices on this list. Voices which said that 
immutability was very useful and could provide optimization 
possibilities. Currently these potential optimizations are unused, 
except when copying the data.

It is very easy to change the design, by changing the single data member 
from type `boost::shared_ptr<const std::string>` to `std::string`. That 
way the small string optimization, if any, of `std::string` could be 
utilized directly.

Come to think of it, the type could actually be specified as a typedef 
of the encoding template parameter.

Regards,
Anders Dalvander

-- 
WWFSMD?