On Wed, Mar 26, 2008 at 07:54:03PM -0400, rgheck wrote:
>> Andre sayeth:
>>
>> On Wed, Mar 26, 2008 at 05:29:22PM -0400, rgheck wrote:
>> > [...]
>> > Comments welcome. And see the long FIXME about remaining issues.
>>
>> > +
>> > + docstring embed(getParam("embed"));
>>
>> We use
>>
>> docstring embed = getParam("embed");
>>
>> [Several occurences]
> If you say so, OK. But there's a difference here, right? And isn't the
> former more efficient?
>
> I ask seriously. That's what I had thought. And you know better than me.
The interesting case is actually
std::string("hello") vs std::string s = "hello";
Conceptionally the former is direct construction whereas the latter is
formally construction of a temporary followed by copy-construction.
So you are right that the latter has the potential of being less
efficient. However, the compiler is free _not_ to perform the copy (even
if there were side effects...) - and that's what pretty much every
compiler I used during the last ten years does.
So even without any optimization formally switched on, both should
produce identical code, and given that, I prefer the latter as it
is one level less of parantheses. A related benefit is that the latter
can't be mistaken as a function declaration whereas the former
sometimes can, and the compiler is required by the Standard authors
to chose the "wrong" option...
Andre'
