www.digitalmars.com         C & C++   DMDScript  

D - D and covariant assignment

reply Friedrich Dominicus <frido q-software-solutions.com> writes:
As I understand this will work
class A
{
	public int foo () {return 0;}
}

class B: A
	public int foo(){ return 1;}
}


A a = new B();
a.foo -> yields 1

But that won't work
class A {
     public int foo () { return 0; }
}

class B: A{
     public int foo () { return 1; }
     public int bar() { return 0;}
}

int main (){
     A a = new B();
     printf("a.foo() = %d\n", a.foo());
     printf("a.bar() = %d\n", a.bar());
     return 0;
}

So D allows sort of covariant assignment, but for accessing a functoin 
defined in a descendant the function to be accessed must be part
of the ancestor too. In the second example bar must be defined in A too.

Is that correct?

Regards
Friedrich
Aug 12 2003
next sibling parent "Mike Wynn" <mike.wynn l8night.co.uk> writes:
"Friedrich Dominicus" <frido q-software-solutions.com> wrote in message
news:bhakdt$o1s$1 digitaldaemon.com...

 As I understand this will work
 class A
 {
 public int foo () {return 0;}
 }

 class B: A
 public int foo(){ return 1;}
 }


 A a = new B();
 a.foo -> yields 1

 But that won't work
 class A {
      public int foo () { return 0; }
 }

 class B: A{
      public int foo () { return 1; }
      public int bar() { return 0;}
 }

 int main (){
      A a = new B();
      printf("a.foo() = %d\n", a.foo());
      printf("a.bar() = %d\n", a.bar());
      return 0;
 }

 So D allows sort of covariant assignment, but for accessing a functoin
 defined in a descendant the function to be accessed must be part
 of the ancestor too. In the second example bar must be defined in A too.


yes, D is statically typed (like Java, C++ Delphi etc) unlike Lua, self
.....

so A a;

although a can hold a reference to B or any other sub class of A you have to
cast a to the subclass before using those "messages" or "methods"
i.e.

      printf("a.bar() = %d\n", (cast(B)a).bar());


cast(B)a  will return null if a is not castable to B

class C { .. }
you could not do cast(C)a as there is not way that `a` can be a subclass of
C (single inheritance only)
the types and inheritance tree for types are all determinded at compile time
late binding is performed via a virtual function table (vtbl) so a.foo()
will be converted to evoke a->vtbl[index of method foo]( a, params to foo );
the compiler has to be able to determine the vtbl index at compile time (not
runtime)

you may have
class D : A {
    void ping() { }
    int bar() { return 44; }
}

a could be a B or D; the vtable entry for bar might be 2 (B) 3 (D) etc ...
Aug 12 2003
prev sibling parent reply Ilya Minkov <midiclub 8ung.at> writes:
Friedrich Dominicus wrote:


 So D allows sort of covariant assignment, but for accessing a functoin 
 defined in a descendant the function to be accessed must be part
 of the ancestor too. In the second example bar must be defined in A too.
 
 Is that correct?


These things must be obvious to you! BTW, how's your own "Q" language doing?

In your example "a" is of type "A", which doesn't have a "bar()" defined.

If you think you know that a is actually of type B, you can go ahead and 
do so:

int main (){
     A a = new B();
     printf("a.foo() = %d\n", a.foo());
     printf("a.bar() = %d\n", cast(B) (a.bar()) );
     return 0;
}

cast will reward you with an exception, if you were wrong.

Can you do such acrobacy in Eiffel or Q as you have tried to suggest 
with your example?

-i.
Aug 12 2003
next sibling parent Mark Evans <Mark_member pathlink.com> writes:
Ilya Minkov says...

BTW, how's your own "Q" language doing?


That's one more for the obscure languages list.  The only Q that I had known
about lived here.
http://www.musikwissenschaft.uni-mainz.de/~ag/q/q.php

Friedrich Dominicus' Q language lives here
http://www.q-software-solutions.com/about.shtml

(Almost all single-letter language names are taken, for those contemplating a
new one.)

Mark
Aug 12 2003
prev sibling parent reply Friedrich Dominicus <frido q-software-solutions.com> writes:
Ilya Minkov wrote:

 Friedrich Dominicus wrote:
 

 So D allows sort of covariant assignment, but for accessing a functoin 
 defined in a descendant the function to be accessed must be part
 of the ancestor too. In the second example bar must be defined in A too.

 Is that correct?


 
 
 These things must be obvious to you! BTW, how's your own "Q" language 
 doing?


Well I had a wrong example on my mind. I rewrote the stuff and found the
that D is correct and I was wrong.



 
 Can you do such acrobacy in Eiffel or Q as you have tried to suggest 
 with your example?


Maybe with an revrese assignment... Quite interesting that at least some
remember me talking about Q years ago. Q kinda works, not all too much I 
admit....

Regards
Friedrich
Aug 14 2003
parent "Daniel Yokomiso" <daniel_yokomiso yahoo.com.br> writes:
"Friedrich Dominicus" <frido q-software-solutions.com> escreveu na mensagem
news:bhfhdv$2ikm$1 digitaldaemon.com...

 Ilya Minkov wrote:

 BTW, how's your own "Q" language doing?


 Well I had a wrong example on my mind. I rewrote the stuff and found the
 that D is correct and I was wrong.


 Can you do such acrobacy in Eiffel or Q as you have tried to suggest
 with your example?


 Maybe with an revrese assignment... Quite interesting that at least some
 remember me talking about Q years ago. Q kinda works, not all too much I
 admit....

 Regards
 Friedrich


Hi,

    I'm also interested in Q, it had some nice ideas in it. Perhaps you
could start a open-source project on Q, compiling it to D ;).

    Best regards,
    Daniel Yokomiso.

"Sincerity is the key. If you can fake that, you've got it made."
 - George Burns



---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.510 / Virus Database: 307 - Release Date: 14/8/2003
Aug 22 2003