Is it possible to reimplement new and delete?

[wyn...@gmail.com]

Recently I had an idea to reimplement new as a function or template
using C malloc. An application example is to implement 'new *this'.
I don't use std-library, most of my codes are 'home-made', only few are
using std:: things (std::terminate(), std::numeric_limits, <typeinfo> )
So is it doable? Any sugestion will be appreciated.

[Alf P. Steinbach]

On 10.05.2018 17:02, wyn...@gmail.com wrote:
> Recently I had an idea to reimplement new as a function or template
> using C malloc.

Why would you want to do use malloc for C++ `new`?

> An application example is to implement 'new *this'.

That's not valid C++ syntax.

So, it's not really an example of anything.

> I don't use std-library, most of my codes are 'home-made', only few are
> using std:: things (std::terminate(), std::numeric_limits, <typeinfo> )
> So is it doable?

Yes, you can override the allocation and deallocation functions,
operator new, operator new[], operator delete and operator delete[], in
the global namespace as well as per class.

> Any sugestion will be appreciated.

I would suggest not doing this. ;-)


Cheers!,

- Alf

[Bo Persson]

Before putting too much work into this, you might want to check what
your compiler does by default.

For example, the compiler I'm using does this:


void* __CRTDECL operator new(size_t const size)
{
for (;;)
{
if (void* const block = malloc(size))
{
return block;
}

if (_callnewh(size) == 0)
{
if (size == SIZE_MAX)
{
__scrt_throw_std_bad_array_new_length();
}
else
{
__scrt_throw_std_bad_alloc();
}
}

// The new handler was successful; try to allocate again...
}
}




Bo Persson

[wyn...@gmail.com]

Alf P. Steinbach於 2018年5月10日星期四 UTC+8下午11時12分18秒寫道：

> On 10.05.2018 17:02, wyn...@gmail.com wrote:
> > Recently I had an idea to reimplement new as a function or template
> > using C malloc.
>
> Why would you want to do use malloc for C++ `new`?
>
>
> > An application example is to implement 'new *this'.
>
> That's not valid C++ syntax.
>
> So, it's not really an example of anything.
>

I think the related 'clone()' problem had been an topic long long ago.
Probably it's just a quick though flashed in my brain.

>
> > I don't use std-library, most of my codes are 'home-made', only few are
> > using std:: things (std::terminate(), std::numeric_limits, <typeinfo> )
> > So is it doable?
>
> Yes, you can override the allocation and deallocation functions,
> operator new, operator new[], operator delete and operator delete[], in
> the global namespace as well as per class.
>
>
> > Any sugestion will be appreciated.
>
> I would suggest not doing this. ;-)
>
>
> Cheers!,
>
> - Alf

pretty encouraging!

[wyn...@gmail.com]

Bo Persson於 2018年5月10日星期四 UTC+8下午11時49分51秒寫道：

Thanks for the example, helpful to know C's malloc can implement new.

Probable I had used a wrong word (reimplement new).

[Paavo Helde]

On 11.05.2018 16:44, wyn...@gmail.com wrote:
> Alf P. Steinbach於 2018年5月10日星期四 UTC+8下午11時12分18秒寫道：
>> On 10.05.2018 17:02, wyn...@gmail.com wrote:
>>> Recently I had an idea to reimplement new as a function or template
>>> using C malloc.
>>
>> Why would you want to do use malloc for C++ `new`?
>>
>>
>>> An application example is to implement 'new *this'.
>>
>> That's not valid C++ syntax.
>>
>> So, it's not really an example of anything.
>>
>
> I think the related 'clone()' problem had been an topic long long ago.
> Probably it's just a quick though flashed in my brain.

The cloning issue should become nicely resolved by the metaclasses
proposal, but I'm told this is still far away.

Anyway, I don't see how malloc would help in any way with cloning.

[wyn...@gmail.com]

Paavo Helde於 2018年5月11日星期五 UTC+8下午11時42分32秒寫道：

Good to know this information, many thanks.

So far, I had only *one* case using the clone function, and the need
of that usecase was not strong enough, neither. I might be also
thinking something else about using malloc for operator new, but not
in a hurry. Probably just preparing to face the future C++.