I just read through the specs about operators and found a strangeness in the definition of opCmp:

http://www.digitalmars.com/d/2.0/operatoroverloading.html#compare

Mathematically the following are equivalent
	a < b       <=>       b > a

But the definition seems to swap
	a < b	 into	  b >= a

I have not tested the compiler about this yet...

On Mon, 08 Mar 2010 06:15:16 -0500, Norbert Nemec <Norbert@nemec-online.de> wrote:

> I just read through the specs about operators and found a strangeness in the definition of opCmp:
>
> http://www.digitalmars.com/d/2.0/operatoroverloading.html#compare
>
> Mathematically the following are equivalent
> 	a < b       <=>       b > a
>
> But the definition seems to swap
> 	a < b	 into	  b >= a
>
> I have not tested the compiler about this yet...

From testing (dmd 2.040) it appears that the spec is wrong (the correct thing is done).  I'm amazed this was never caught, I think it's been this way for a long time.

Good catch, you should file a bug.

test code:


import std.stdio;

struct S
{
}

struct T
{
    int opCmp(S s)
    {
        return 0;
    }
}

void main()
{
    S s;
    T t;
    writeln(s < t);
    writeln(t < s);
    writeln(s <= t);
    writeln(t <= s);
    writeln(s > t);
    writeln(t > s);
    writeln(s >= t);
    writeln(t >= s);
}

outputs:

false
false
true
true
false
false
true
true

-Steve