From: Peter Dimov (pdimov_at_[hidden])
Date: 2004-01-22 07:56:59
Daryle Walker wrote:
> On 1/20/04 10:05 PM, "Howard Hinnant" <hinnant_at_[hidden]> wrote:
>
>> But I suggest an alternative interface (with similar functionality):
>>
>> move_ptr<T>
>> move_ptr<T[]>
>>
>> The advantage of my suggestion boils down to nothing more than
>> syntax. There are no generic coding advantages that I am aware of.
>> I simply submit that S<T[]> is slightly more intuitive than
>> S_array<T>.
>
> I think the reverse; that "S<T[]>" is worse, because it's
> anti-intuitive! When I see a specialization, I think of something
> that has the same interface, but with some sort of optimized
> implementation.
That's what I said at the time Howard first proposed ptr<T[]>, too. But...
> Your idea is bad because you're deliberately giving
> this specialization a _different_ interface. It's worse than the
> "std::vector<bool> sucks compared to the unspecialized
> std::vector<T>" quirk (vector's problem is in the subtle difference
> in the dereferencing interfaces).
... this is not true at all. The problem with vector<bool> is that the
primary template generates a perfectly fine vector<bool>, but the
specialization shadows that with something that is incompatible (not a
vector<>, not even a standard container w.r.t. container requirements.)
ptr<T[]>, as generated by the primary template, doesn't make sense. If your
intuition tells you that it has the same interface, your intuition is wrong,
because it can't.