To me it sounds more like you should have a "objects are alive when
allocated"-vector, so that you construct all objects when you allocate your
data and just "assigns" them during push-backs and the like. But I don't
think that should be part of the standard either. It is relatively easy to
implement yourself.

// Robin

On Sat, Dec 28, 2024, 09:14 Jonathan Wakely via Std-Proposals <
std-proposals_at_[hidden]> wrote:

>
>
> On Sat, 28 Dec 2024, 00:37 Magnus Fromreide, <magfr_at_[hidden]> wrote:
>
>> On Sat, Dec 28, 2024 at 12:10:55AM +0000, Jonathan Wakely via
>> Std-Proposals wrote:
>> > On Fri, 27 Dec 2024, 22:51 Chris Ryan via Std-Proposals, <
>> > std-proposals_at_[hidden]> wrote:
>> >
>> > > Hi Andre,
>> > > .resize(0) actually deletes the storage.
>> > >
>> >
>> > No it doesn't. clear() and resize(0) both preserve capacity.
>> >
>> >
>> > Farrakh is wanting it to preserve the storage (keep the .capacity(), not
>> > > have to alloc it again)
>> > >
>> >
>> > That's what clear() does.
>> >
>> > The only reason to add a new function would be to avoid the destructor
>> > calls for the elements, which can be accomplished with a custom
>> Allocator
>> > that has an empty destroy function.
>>
>> Wouldn't that lead to the member objects not beeing destructed when the
>> vector's lifetime ends?
>>
>
> Yes, because I thought that was the point. I'm not saying I support the
> idea. I certainly don't want it added as a member function of std::vector.
>
> If you want to manage object lifetime incorrectly, you should do that
> yourself without help from the standard.
>
>
>
>
>
>> To me it sounds like Farrakh wants an extra abstraction on top of vector
>> which
>> maintains it's own size which is <= vector.size.
>>
>> This should be possible to do using a decorator like
>>
>> template<class C>
>> class Farrakh {
>> private:
>> C c;
>> typename C::size_type size_ = 0;
>> public:
>> // lots of members that forward to c but uses size_ rather than
>> // c.size(), but in particular these:
>>
>> void push_back(auto v) {
>> if (size_ < c.size())
>> c[size_++] = v;
>> else
>> c.push_back(v);
>> }
>> void clear() { size_ = 0; }
>> };
>>
>> I am strongly against anything like this beeing added to the standard.
>>
>> >
>> >
>> > > Farrakh,
>> > > I think that you might not realize the reason .clear() is slower is
>> > > because it has to call the in-place destructor for each element (zero
>> > > through size(), but not all the way up to .capacity()) You can't just
>> > > change the count or you screw up the object lifetime of those objects
>> > > that were not destructed. D'tors won't get called. Index positions
>> may
>> > > potentially get overwritten later with subsequent push_backs(...) and
>> the
>> > > in-place c'tors calls. (thus causing leaks)
>> > >
>> > > https://en.cppreference.com/w/cpp/container/vector/clear
>> > >
>> > >
>> > > On Fri, Dec 27, 2024 at 2:08 PM Andre Kostur via Std-Proposals <
>> > > std-proposals_at_[hidden]> wrote:
>> > >
>> > >> What’s wrong with .resize(0) ?
>> > >>
>> > >> On Fri, Dec 27, 2024 at 1:57 PM Фаррах Фаттахов via Std-Proposals <
>> > >> std-proposals_at_[hidden]> wrote:
>> > >>
>> > >>> After working with SFML, I so hated std::vector (I was a bad
>> programmer
>> > >>> back then and didn't know about the reserve() function, which led
>> to not
>> > >>> understanding why std::vector is so slow, while a simple array
>> performs
>> > >>> much better). As a result, in my new project i created my own
>> version of
>> > >>> std::vector - FE2D::dynamic_array. And there, from the beginning i
>> created
>> > >>> one useful function - reset()
>> > >>> ```C++
>> > >>>
>> > >>> // Reset the Size of Dynamic Array to Zero, but it's don't Touches
>> the Real Occupied Memory of Dynamic Array
>> > >>> void reset() { mSize = 0; }
>> > >>> ```
>> > >>>
>> > >>> Instead of clear() it's just setting mSize to zero. It's very
>> useful in
>> > >>> the situations like
>> > >>>
>> > >>> ```C++
>> > >>> void Render() { // It's called every frame
>> > >>> // Add a sprites to the FE2D::dynamic_array to draw it later
>> > >>> m_Renderer.AddSprite(m_Sprite_0);
>> > >>> m_Renderer.AddSprite(m_Sprite_1);
>> > >>> m_Renderer.AddSprite(m_Sprite_2);
>> > >>> ...
>> > >>> // Draw the Sprites and reset the dynamic arrays to prepare it for
>> the next frame
>> > >>> m_Renderer.RenderSprites();
>> > >>> }
>> > >>> ```
>> > >>> Here reset() works like clear(), but much faster
>> > >>>
>> > >>> In the std::vector or std::deque is no functions like reset() and i
>> want
>> > >>> to add it there, especially sometimes i get errors because my
>> > >>> FE2D::dynamic_array is not so good like std::vector and i have to
>> use it
>> > >>> instead of mine, but there is no function i need. I think reset()
>> will be
>> > >>> nice thing for std::vector.
>> > >>> --
>> > >>> Std-Proposals mailing list
>> > >>> Std-Proposals_at_[hidden]
>> > >>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>> > >>>
>> > >> --
>> > >> Std-Proposals mailing list
>> > >> Std-Proposals_at_[hidden]
>> > >> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>> > >>
>> > > --
>> > > Std-Proposals mailing list
>> > > Std-Proposals_at_[hidden]
>> > > https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>> > >
>>
>> > --
>> > Std-Proposals mailing list
>> > Std-Proposals_at_[hidden]
>> > https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>>
>> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>