Overloading float operators

[ag0aep6g]

On 12/11/2017 08:28 PM, rumbu wrote:
> Is there any way to overload specific floating point operators?
> https://dlang.org/spec/expression.html#floating-point-comparisons

Those don't seem to work anymore. At least since 2.073, dmd rejects them 
and says to use std.math.isNaN instead. Looks like someone forgot to 
upate the spec. I couldn't find anything in the changelog either.

> I'm using a decimal data type (a struct) and one of the possible values 
> is NaN, that's why I need these operators.

You can return float.nan from opCmp to mean "unordered":

----
struct S
{
     bool isNaN = true;
     int value;

     this(int value) { this.isNaN = false; this.value = value; }

     bool opEquals(const S other) const pure nothrow @safe @nogc
     {
         return !this.isNaN && !other.isNaN && this.value == other.value;
     }

     float opCmp(const S other) const pure nothrow @safe @nogc
     {
         if (this.isNaN || other.isNaN) return float.nan;
         if (this.value < other.value) return -1;
         if (this.value > other.value) return 1;
         return 0;
     }
}

void main()
{
     S s1; // NaN
     S s2; // NaN

     assert(s1 != s2); // neither equal ...
     assert(!(s1 < s2)); // nor less than ...
     assert(!(s1 > s2)); // nor greater
}
----

> I know also that this also was discussed, but is there any way to 
> separately implement == and !=, so both return true or false in the same 
> time?

I don't think so.

> The reason is the same: NaN == NaN = false and NaN != NaN = false in the 
> same time.

But NaN != NaN is true.