Re: [boost] Re: container design (Was: std::complex design (Was: N-Dimensional array))

From: nbecker_at_[hidden]
Date: 2002-03-08 15:33:50

>>>>> "rwgk" == rwgk  <rwgk_at_[hidden]> writes:

    rwgk> --- In boost_at_y..., "Peter Dimov" <pdimov_at_m...> wrote:
    >> From: "Howard Hinnant" <hinnant_at_t...>
    >> > Meanwhile vector<double> v(10, uninitialized_flag) breaks no 
    rwgk> vector
    >> > invariants.  The vector is destructible.  The vector elements can 
    rwgk> be
    >> > assigned into.  If your hardware doesn't mind (mine doesn't), you 
    rwgk> can
    >> > even assign from this uninitialized vector.  I  can't think of a 
    rwgk> reason
    >> > why you would want to do that though.
    >> 
    >> You can't v.insert(v.begin(), 5.0) because copying the 
    rwgk> uninitialized doubles
    >> is undefined behavior. The insert() preconditions do not say that 
    rwgk> you can't.
    >> Something is wrong. :-)

    rwgk> In my mind

    rwgk> vector<double> vec(100000, uninitialized_flag());

    rwgk> means these things:

    rwgk> 1. I do not want to waste time for something that I am
    rwgk> absolutely sure I do not need (the default construction)
    rwgk> because

    rwgk> 2. I am about to fill the memory with the given number
    rwgk> of elements by some other facility, and

    rwgk> 3. I am taking responsibility and I have read the
    rwgk> documentation for uninitialized_flag() which told me
    rwgk> to initialize the memory before doing /anything/ else
    rwgk> with it.

    rwgk> Being safe by default is great, but IMO forcing everyone to
    rwgk> waste time only to be safe under any conceivable eventuality
    rwgk> is misguided and may be one of the reasons why C++ still has
    rwgk> a bad reputation among many (non-computer) scientists.

I completely agree.  That's one reason I have not really understood
all of the conversation on this subject (the other being that some of
it is over my head).  I get the impression that much discussion has
been on how to prevent undefined things from happening if I use
unitialized memory.  I was under the impression that I couldn't care
less what happens.  That's a programming error.  Too bad.  If it
segfaults, that's a good thing.

Or maybe I misunderstood.