Re: [Boost-users] [scoped_array] "optimization"

From: Michael Marcin (mmarcin_at_[hidden])
Date: 2007-04-25 19:42:22

David Daeschler wrote:
>
> Straight out of the docs
> http://www.boost.org/libs/smart_ptr/scoped_array.htm
>
>> Because scoped_array is so simple, in its usual implementation every
>> operation is as fast as a built-in array pointer and it has no more
>> space overhead that a built-in array pointer.
>
> IMHO saying that the change to an array pointer is an "optimization"
> at all is dubious.
>

That's pretty much what I figured.  Also I believe the added inline keyword 
is redundant as a function defined inside of a class declaration is the same 
as defining inline after the closing semi-colon of the class correct?

I originally wrote that using a scoped_array to document the lifetime, 
ownership, and type of data pointed to.  And so I didn't have to worry about 
things like forgetting to call delete[] instead of delete.  I figured that 
any compiler worth using should inline this out to the same code as the 
"optimized" version but I have no proof of that I was hoping someone would 
have done simple tests comparing generated code that I could reference when 
I discuss these "improvements" with the team later this week.

The copy constructor argument Richard Hadsell mentioned I hadn't even 
considered and it is a good argument in its own right.

What are the likely results of calling operator delete instead of operator 
delete[] on an array of builtins allocated with operator new[]?  Is it 
undefined behavior?

Thanks,

Michael Marcin