In some code I'm working on, my asserted when I've confirmed it was correct. With the opCmp() overridden.  'this' refers to the current object, so why doesn't the first one succeed?

class X {
  int z;

  this(int xx) {
	z = xx;
  }

  override int opCmp(Object y){
	X x = cast(X) y;
	return z - x.z;
  }

  void something(X rhs) {
    assert(this.opCmp(rhs) == 0); //works fine.
    assert(this == rhs); 	  //fails?
  }
}

int main(){
	X a = new X(10);
	X b = new X(10); //different objects, same compare value
	a.something(b);
	return 0;
}

On Monday, April 30, 2012 07:09:46 Era Scarecrow wrote:
>   In some code I'm working on, my asserted when I've confirmed it
> was correct. With the opCmp() overridden.  'this' refers to the
> current object, so why doesn't the first one succeed?
> 
> class X {
>    int z;
> 
>    this(int xx) {
> 	z = xx;
>    }
> 
>    override int opCmp(Object y){
> 	X x = cast(X) y;
> 	return z - x.z;
>    }
> 
>    void something(X rhs) {
>      assert(this.opCmp(rhs) == 0); //works fine.
>      assert(this == rhs); 	  //fails?
>    }
> }
> 
> int main(){
> 	X a = new X(10);
> 	X b = new X(10); //different objects, same compare value
> 	a.something(b);
> 	return 0;
> }

== uses opEquals, not opCmp. It's using Object's opEquals, which does a comparison of the references, so it's true when comparing the exact same object and false otherwise.

- Jonathan M Davis

On Monday, 30 April 2012 at 05:22:56 UTC, Jonathan M Davis wrote:
>
> == uses opEquals, not opCmp. It's using Object's opEquals, which does a
> comparison of the references, so it's true when comparing the exact same
> object and false otherwise.

 Ahhh of course. Personally I think opCmp includes opEquals... At least I only have to deal with 2 compare functions and not more.

On Monday, April 30, 2012 07:28:16 Era Scarecrow wrote:
> On Monday, 30 April 2012 at 05:22:56 UTC, Jonathan M Davis wrote:
> > == uses opEquals, not opCmp. It's using Object's opEquals,
> > which does a
> > comparison of the references, so it's true when comparing the
> > exact same
> > object and false otherwise.
> 
>   Ahhh of course. Personally I think opCmp includes opEquals... At
> least I only have to deal with 2 compare functions and not more.

I believe that there are two major reasons why opEquals is separate:

1. Many types of objects can have equality but less than and greater than comparisons would make no sense for them.

2. It's more efficient to check for equality with opEquals than it would be with opCmp.

- Jonathan M Davis