I can't find a way to get a class sizeof property - it returns 4 (32bit pointer size) always. I tried many ways but still can't figure out how to do this. Documentation says that

".sizeof applied to a class object returns the size of the class reference, not the class instantiation."

Is it possible? Any idea is much appreciated.

Denis Koroskin wrote:
> I can't find a way to get a class sizeof property - it returns 4 (32bit pointer size) always. I tried many ways but still can't figure out how to do this. Documentation says that
> 
> ".sizeof applied to a class object returns the size of the class reference, not the class instantiation."
> 
> Is it possible? Any idea is much appreciated.

No nice way to do it..

auto sz = myclass.classinfo.init.length;

problem this is not a compiletime constant,

In D2 you can use __traits to get it...

Kinda lame if you ask me :P

how about .instancesizeof ? I doubt that's even a remotely common variable name ...

-Tomas

On Fri, 14 Nov 2008 21:06:39 +0300, Tomas Lindquist Olsen <tomas@famolsen.dk> wrote:

> Denis Koroskin wrote:
>> I can't find a way to get a class sizeof property - it returns 4 (32bit pointer size) always. I tried many ways but still can't figure out how to do this. Documentation says that
>>  ".sizeof applied to a class object returns the size of the class reference, not the class instantiation."
>>  Is it possible? Any idea is much appreciated.
>
> No nice way to do it..
>
> auto sz = myclass.classinfo.init.length;
>
> problem this is not a compiletime constant,
>
> In D2 you can use __traits to get it...
>
> Kinda lame if you ask me :P
>
> how about .instancesizeof ? I doubt that's even a remotely common variable name ...
>
> -Tomas

__traits trick works nice, thank you!

BTW, I found small inconsistency/bug:

template SizeOfPointee(T)
{
    static if (is (T t)) {
        enum SizeOfPointee = typeof(*t).sizeof;
    }
}

template SizeOfPointer(T)
{
    static if (is (T t)) {
        enum SizeOfPointer = typeof(t).sizeof;
    }
}

int i = SizeOfPointee!(int*); // works ok
int j = SizeOfPointer!(int*); // fails

I believe both should work.

Tomas Lindquist Olsen wrote:
> Denis Koroskin wrote:
>> I can't find a way to get a class sizeof property - it returns 4 (32bit pointer size) always. I tried many ways but still can't figure out how to do this. Documentation says that
>>
>> ".sizeof applied to a class object returns the size of the class reference, not the class instantiation."
>>
>> Is it possible? Any idea is much appreciated.
> 
> No nice way to do it..
> 
> auto sz = myclass.classinfo.init.length;
> 
> problem this is not a compiletime constant,
> 
> In D2 you can use __traits to get it...
> 
> Kinda lame if you ask me :P
> 
> how about .instancesizeof ? I doubt that's even a remotely common variable name ...

I added an "isizeof" proposal to the issue tracker maybe two years ago.  Feel free to vote it up if you'd like.


Sean

On Fri, Nov 14, 2008 at 2:35 PM, Sean Kelly <sean@invisibleduck.org> wrote:
> Tomas Lindquist Olsen wrote:
>>
>> Denis Koroskin wrote:
>>>
>>> I can't find a way to get a class sizeof property - it returns 4 (32bit pointer size) always. I tried many ways but still can't figure out how to do this. Documentation says that
>>>
>>> ".sizeof applied to a class object returns the size of the class reference, not the class instantiation."
>>>
>>> Is it possible? Any idea is much appreciated.
>>
>> No nice way to do it..
>>
>> auto sz = myclass.classinfo.init.length;
>>
>> problem this is not a compiletime constant,
>>
>> In D2 you can use __traits to get it...
>>
>> Kinda lame if you ask me :P
>>
>> how about .instancesizeof ? I doubt that's even a remotely common variable name ...
>
> I added an "isizeof" proposal to the issue tracker maybe two years ago.
>  Feel free to vote it up if you'd like.

But W will say "it's already accessible through __traits."  Not to sound fatalistic, but do you honestly think this would be added to D1?

A few years too late to by why isn't all classes objects, an instance of a special virtual class called Class.

On Sat, 15 Nov 2008 08:55:29 +1300, Jarrett Billingsley <jarrett.billingsley@gmail.com> wrote:

> On Fri, Nov 14, 2008 at 2:35 PM, Sean Kelly <sean@invisibleduck.org> wrote:
>> Tomas Lindquist Olsen wrote:
>>>
>>> Denis Koroskin wrote:
>>>>
>>>> I can't find a way to get a class sizeof property - it returns 4 (32bit
>>>> pointer size) always. I tried many ways but still can't figure out how to do
>>>> this. Documentation says that
>>>>
>>>> ".sizeof applied to a class object returns the size of the class
>>>> reference, not the class instantiation."
>>>>
>>>> Is it possible? Any idea is much appreciated.
>>>
>>> No nice way to do it..
>>>
>>> auto sz = myclass.classinfo.init.length;
>>>
>>> problem this is not a compiletime constant,
>>>
>>> In D2 you can use __traits to get it...
>>>
>>> Kinda lame if you ask me :P
>>>
>>> how about .instancesizeof ? I doubt that's even a remotely common variable
>>> name ...
>>
>> I added an "isizeof" proposal to the issue tracker maybe two years ago.
>>  Feel free to vote it up if you'd like.
>
> But W will say "it's already accessible through __traits."  Not to
> sound fatalistic, but do you honestly think this would be added to D1?

Jarrett Billingsley:
> But W will say "it's already accessible through __traits."  Not to sound fatalistic, but do you honestly think this would be added to D1?

Generally I think that having a fully frozen language (D1), that can't accept improvements, is bad.
And I think that in an open source language, if enough people what to add a feature or remove a bug, and they are able to implement/do it, then it has to be doable. Begging a single person to implement all the changes or all debugging seems silly. If enough people are willing, then such feature/debugging can be added/done to LDC, for example.

Bye,
bearophile

Jarrett Billingsley wrote:
> On Fri, Nov 14, 2008 at 2:35 PM, Sean Kelly <sean@invisibleduck.org> wrote:
>>
>> I added an "isizeof" proposal to the issue tracker maybe two years ago.
>>  Feel free to vote it up if you'd like.
> 
> But W will say "it's already accessible through __traits."  Not to
> sound fatalistic, but do you honestly think this would be added to D1?

I don't :-)  If it were, it would have been added already.


Sean