On 05/30/2013 03:02 PM, Joseph Rushton Wakeling wrote:
> I would never have thought of that trick with the second template
parameter.
Phobos is a source of ideas. ;)
> I guess it might fall over with complex structs or classes, though. :-\
Copying structs is trivial because they already have copy semantics:
T b = a; // b is a copy of a
However, that depends on correctly implemented copy semantics on T.
For classes, there is no syntax for copying. The type may have annotated
a member function that it is the duplication function or we may know by
convention that dup() is the equivalent of array .dup.
> What do you reckon the impact of this will be performance-wise?
As long there is no extra copy generated during the process, it should
be fine. However, the usual issues around the current conservative GC
applies. :/
Ali
