www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - cast(size_t)&c != 0, but c is null

reply "simendsjo" <simendsjo gmail.com> writes:
Very newbie question coming up :)

How does D mark null values for classes?
`c is null` returns true, but `&c` isn't 0.
So how does D know `c is null`?

class C {}
C c;
assert(c is null);
assert(cast(size_t)&c == 0); // fails.
May 14 2013
next sibling parent reply "Andrej Mitrovic" <andrej.mitrovich gmail.com> writes:
On Tuesday, 14 May 2013 at 12:18:20 UTC, simendsjo wrote:

 Very newbie question coming up :)

 How does D mark null values for classes?
 `c is null` returns true, but `&c` isn't 0.
 So how does D know `c is null`?

 class C {}
 C c;
 assert(c is null);
 assert(cast(size_t)&c == 0); // fails.


That's the address of the reference. If you want the address of 
the object:

assert(cast(size_t)cast(void*)c == 0); // passes
May 14 2013
parent "simendsjo" <simendsjo gmail.com> writes:
On Tuesday, 14 May 2013 at 12:20:00 UTC, Andrej Mitrovic wrote:

 On Tuesday, 14 May 2013 at 12:18:20 UTC, simendsjo wrote:

 Very newbie question coming up :)

 How does D mark null values for classes?
 `c is null` returns true, but `&c` isn't 0.
 So how does D know `c is null`?

 class C {}
 C c;
 assert(c is null);
 assert(cast(size_t)&c == 0); // fails.


 That's the address of the reference. If you want the address of 
 the object:

 assert(cast(size_t)cast(void*)c == 0); // passes


Ah. Thanks to you both. Now I have some strange bug where 
somewhere in my code (probably using alias for passing symbols to 
templates), it get's a "wrong" reference...
May 14 2013
prev sibling parent reply "Dicebot" <m.strashun gmail.com> writes:
On Tuesday, 14 May 2013 at 12:18:20 UTC, simendsjo wrote:

 Very newbie question coming up :)

 How does D mark null values for classes?
 `c is null` returns true, but `&c` isn't 0.
 So how does D know `c is null`?

 class C {}
 C c;
 assert(c is null);
 assert(cast(size_t)&c == 0); // fails.


"&c" is address of reference, no class instance. I don't know if 
there is a way to get a pointer to class instance in D but I am 
not aware of one.
May 14 2013
parent reply David <d dav1d.de> writes:
Am 14.05.2013 14:24, schrieb Dicebot:

 On Tuesday, 14 May 2013 at 12:18:20 UTC, simendsjo wrote:

 Very newbie question coming up :)

 How does D mark null values for classes?
 `c is null` returns true, but `&c` isn't 0.
 So how does D know `c is null`?

 class C {}
 C c;
 assert(c is null);
 assert(cast(size_t)&c == 0); // fails.


 
 "&c" is address of reference, no class instance. I don't know if there
 is a way to get a pointer to class instance in D but I am not aware of one.


A simple cast to void* should do it: cast(void*)c
May 15 2013
parent reply "Dicebot" <m.strashun gmail.com> writes:
On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:

 "&c" is address of reference, no class instance. I don't know 
 if there
 is a way to get a pointer to class instance in D but I am not 
 aware of one.


 A simple cast to void* should do it: cast(void*)c


Is it actually defined somewhere in spec?
May 15 2013
parent reply Artur Skawina <art.08.09 gmail.com> writes:
On 05/15/13 13:04, Dicebot wrote:

 On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:

 "&c" is address of reference, no class instance. I don't know if there
 is a way to get a pointer to class instance in D but I am not aware of one.


 A simple cast to void* should do it: cast(void*)c




   class C { auto opCast(T:void*)() { return null; } }

So - no - a simple cast to void* won't always work, as the op can be
overloaded, even if only by accident. 

See also:
http://forum.dlang.org/thread/mutvhrrdfzunbrhmvztp forum.dlang.org#post-mailman.2432.1354406246.5162.digitalmars-d-learn:40puremagic.com
and
http://d.puremagic.com/issues/show_bug.cgi?id=8545


artur
May 15 2013
next sibling parent reply "Steven Schveighoffer" <schveiguy yahoo.com> writes:
On Wed, 15 May 2013 11:08:33 -0400, Artur Skawina <art.08.09 gmail.com>  
wrote:


 On 05/15/13 13:04, Dicebot wrote:

 On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:

 "&c" is address of reference, no class instance. I don't know if there
 is a way to get a pointer to class instance in D but I am not aware  
 of one.


 A simple cast to void* should do it: cast(void*)c




    class C { auto opCast(T:void*)() { return null; } }

 So - no - a simple cast to void* won't always work, as the op can be
 overloaded, even if only by accident.


You can insert a cast to Object first.

-Steve
May 15 2013
parent reply Timothee Cour <thelastmammoth gmail.com> writes:
 You can insert a cast to Object first.




doesn't seem to work.
I'm trying to write a generic solution but having an innocent 'T opCast(T :
int)() { return 1; }' breaks it, see Error below.
doesn't seem to work.

This seems to call for a compiler solution?


 import std.stdio;
import std.conv;


void*AddressOf(T)(T a) if(is(T==class)){
    return cast(void*)cast(Object) a;//Error: template instance opCast!(
Object) opCast!(Object) does not match template declaration opCast(T : int)(
)
}
T*AddressOf(T)(ref T a) if(!is(T==class)){
    return &a;
}
class A{
    int x;
    T opCast(T : int)() { return 1; }
//    int opCast(T:void*)() { return 1; } //won't work either
}
struct B{

}
void main(){
    A a;
    writeln(AddressOf(a));
    a = new A;
    writeln(AddressOf(a));

    B b;
    writeln(AddressOf(b));
    B*b2;
    writeln(AddressOf(*b2));
//    writeln(AddressOf(B.init)); //waiting for rvalue ref, DIP39
}


On Wed, May 15, 2013 at 11:30 AM, Steven Schveighoffer
<schveiguy yahoo.com>wrote:


 On Wed, 15 May 2013 11:08:33 -0400, Artur Skawina <art.08.09 gmail.com>
 wrote:

  On 05/15/13 13:04, Dicebot wrote:

 On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:


 "&c" is address of reference, no class instance. I don't know if there

 is a way to get a pointer to class instance in D but I am not aware of
 one.


 A simple cast to void* should do it: cast(void*)c




    class C { auto opCast(T:void*)() { return null; } }

 So - no - a simple cast to void* won't always work, as the op can be
 overloaded, even if only by accident.


 You can insert a cast to Object first.

 -Steve
May 15 2013
next sibling parent reply Dmitry Olshansky <dmitry.olsh gmail.com> writes:
15-May-2013 23:17, Timothee Cour пишет:

  >> You can insert a cast to Object first.

 doesn't seem to work.
 I'm trying to write a generic solution but having an innocent
 'TopCast(T:int)(){return1;}' breaks it, see Error below.
 doesn't seem to work.

 This seems to call for a compiler solution?


Won't this work?
*cast(void**)&object

 importstd.stdio;
 importstd.conv;


 void*AddressOf(T)(Ta)if(is(T==class)){
 returncast(void*)cast(Object)a;//Error:templateinstanceopCast!(Object)opCast!(Object)doesnotmatchtemplatedeclarationopCast(T:int)()
 }
 T*AddressOf(T)(refTa)if(!is(T==class)){
 return&a;
 }
 classA{
 intx;
 TopCast(T:int)(){return1;}
 //intopCast(T:void*)(){return1;} //won't work either
 }
 structB{

 }
 voidmain(){
 Aa;
 writeln(AddressOf(a));
 a=newA;
 writeln(AddressOf(a));

 Bb;
 writeln(AddressOf(b));
 B*b2;
 writeln(AddressOf(*b2));
 //writeln(AddressOf(B.init));//waitingforrvalueref,DIP39
 }


 On Wed, May 15, 2013 at 11:30 AM, Steven Schveighoffer
 <schveiguy yahoo.com <mailto:schveiguy yahoo.com>> wrote:

     On Wed, 15 May 2013 11:08:33 -0400, Artur Skawina
     <art.08.09 gmail.com <mailto:art.08.09 gmail.com>> wrote:

         On 05/15/13 13:04, Dicebot wrote:

             On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:

                     "&c" is address of reference, no class instance. I
                     don't know if there
                     is a way to get a pointer to class instance in D but
                     I am not aware of one.


                 A simple cast to void* should do it: cast(void*)c


             class C { auto opCast(T:void*)() { return null; } }

         So - no - a simple cast to void* won't always work, as the op can be
         overloaded, even if only by accident.


     You can insert a cast to Object first.

     -Steve



-- 
Dmitry Olshansky
May 15 2013
next sibling parent Timothee Cour <thelastmammoth gmail.com> writes:
 Won't this work?
 *cast(void**)&object





great, works!

I've added it to my dtools repo:

https://github.com/timotheecour/dtools/blob/master/dtools/util/util.d#L19



 May 15 2013



prev sibling  parent reply "Dicebot" <m.strashun gmail.com>  writes:


On Wednesday, 15 May 2013 at 19:30:41 UTC, Dmitry Olshansky wrote:

 Won't this work?
 *cast(void**)&object


Right question is "is this guaranteed to work by spec?".



 May 16 2013



  parent reply "Maxim Fomin" <maxim maxim-fomin.ru>  writes:


On Thursday, 16 May 2013 at 09:05:46 UTC, Dicebot wrote:

 On Wednesday, 15 May 2013 at 19:30:41 UTC, Dmitry Olshansky 
 wrote:

 Won't this work?
 *cast(void**)&object


 Right question is "is this guaranteed to work by spec?".


Spec is silent about this but in practice this should work. Real 
question is here why would you want "pointer to class" when you 
have normal reference and what such "pointer" would give you.



 May 16 2013



 next sibling parent  Timothee Cour <thelastmammoth gmail.com>  writes:


interface with C++ for example (for example extern(C++) classes )

On Thu, May 16, 2013 at 2:31 AM, Maxim Fomin <maxim maxim-fomin.ru> wrote:


 On Thursday, 16 May 2013 at 09:05:46 UTC, Dicebot wrote:


 On Wednesday, 15 May 2013 at 19:30:41 UTC, Dmitry Olshansky wrote:


 Won't this work?
 *cast(void**)&object


 Right question is "is this guaranteed to work by spec?".


 Spec is silent about this but in practice this should work. Real question
 is here why would you want "pointer to class" when you have normal
 reference and what such "pointer" would give you.





 May 16 2013



prev sibling next sibling parent  "Dicebot" <m.strashun gmail.com>  writes:


On Thursday, 16 May 2013 at 09:31:11 UTC, Maxim Fomin wrote:

 On Thursday, 16 May 2013 at 09:05:46 UTC, Dicebot wrote:

 On Wednesday, 15 May 2013 at 19:30:41 UTC, Dmitry Olshansky 
 wrote:

 Won't this work?
 *cast(void**)&object


 Right question is "is this guaranteed to work by spec?".


 Spec is silent about this but in practice this should work. 
 Real question is here why would you want "pointer to class" 
 when you have normal reference and what such "pointer" would 
 give you.


Then spec needs to be updated or this is not correct solution.



 May 16 2013



prev sibling  parent  Dmitry Olshansky <dmitry.olsh gmail.com>  writes:


16-May-2013 13:31, Maxim Fomin пишет:

 On Thursday, 16 May 2013 at 09:05:46 UTC, Dicebot wrote:

 On Wednesday, 15 May 2013 at 19:30:41 UTC, Dmitry Olshansky wrote:

 Won't this work?
 *cast(void**)&object


 Right question is "is this guaranteed to work by spec?".


 Spec is silent about this but in practice this should work. Real
 question is here why would you want "pointer to class" when you have
 normal reference and what such "pointer" would give you.


Since class is a reference it's a pointer value (address anyhow).
Hence you'll get an address of value holding that address (e.g. of a 
temporary) then dereference it as void** pointer.

It's sometimes useful in this particular case to circumvent type-system 
in order to do smth (interface with C etc.).



-- 
Dmitry Olshansky



 May 16 2013



prev sibling  parent reply "Namespace" <rswhite4 googlemail.com>  writes:


 //    writeln(AddressOf(B.init)); //waiting for rvalue ref, 
 DIP39


Did I miss something? Since when is sure that the DIP is 
implemented? o.O



 May 15 2013



  parent reply Timothee Cour <thelastmammoth gmail.com>  writes:


On Wed, May 15, 2013 at 12:35 PM, Namespace <rswhite4 googlemail.com> wrote:


 //    writeln(AddressOf(B.init)); //waiting for rvalue ref, DIP39


 Did I miss something? Since when is sure that the DIP is implemented? o.O


just wishful thinking :) actually this should've been:
//    writeln(AddressOf(B.init^)); //waiting for rvalue ref, DIP39



 May 15 2013



  parent  "Namespace" <rswhite4 googlemail.com>  writes:


On Wednesday, 15 May 2013 at 20:15:26 UTC, Timothee Cour wrote:

 On Wed, May 15, 2013 at 12:35 PM, Namespace 
 <rswhite4 googlemail.com> wrote:


 //    writeln(AddressOf(B.init)); //waiting for rvalue ref, 
 DIP39


 Did I miss something? Since when is sure that the DIP is 
 implemented? o.O


 just wishful thinking :) actually this should've been:
 //    writeln(AddressOf(B.init^)); //waiting for rvalue ref, 
 DIP39


I know that feeling. :D But I doubt that D will have such a thing 
some day.
But: good luck, it is about time that D catch up fully to C++. :)



 May 15 2013



prev sibling  parent  David <d dav1d.de>  writes:


Am 15.05.2013 17:08, schrieb Artur Skawina:

 On 05/15/13 13:04, Dicebot wrote:

 On Wednesday, 15 May 2013 at 10:31:29 UTC, David wrote:

 "&c" is address of reference, no class instance. I don't know if there
 is a way to get a pointer to class instance in D but I am not aware of one.


 A simple cast to void* should do it: cast(void*)c




 
    class C { auto opCast(T:void*)() { return null; } }
 
 So - no - a simple cast to void* won't always work, as the op can be
 overloaded, even if only by accident. 
 
 See also:
 http://forum.dlang.org/thread/mutvhrrdfzunbrhmvztp forum.dlang.org#post-mailman.2432.1354406246.5162.digitalmars-d-learn:40puremagic.com
 and
 http://d.puremagic.com/issues/show_bug.cgi?id=8545
 
 
 artur
 


Overloading cast is broken and shouldn't be in the language (imo). Why
broken? If you only want to overload cast(bool) (for if(x) which calls
if(cast(bool)x -> x.opCast!bool()) you have to overload all other
possible casts, otherwise compiler will complain that this cast is no
overloaded. Also overloading a cast doesn't really make any sense
(except the bool case), that's also the reason why I really hate
std.conv.to's new behaviour of calling opCast, this makes no sense, to!
does a conversion not a cast.



 May 16 2013