Issue 14382 - converting old D1 operator overloading style to new D2 style

[schuetzm]

Patches by ketmar: https://issues.dlang.org/show_bug.cgi?id=14382

[Safety0ff]

Previous discussions for reference:
#2093
http://forum.dlang.org/thread/ifghzjafvfqrqkhlpinc@forum.dlang.org
https://issues.dlang.org/show_bug.cgi?id=12462

[schveiguy]

All these functions are nearly identical. The whole point of having the template was so you can easily implement multiple operators with one set of code. For example:

VariantN opBinary(string op, T)(T rhs) if (op == "+" || op == "-" || op = "*" || op = "/")
{
    return opArithmetic!(T, op)(rhs);
}

In fact, I'd say get rid of opArithmetic and opLogic, and just move the implementations into an appropriate opBinary

[yebblies]

This makes the code less readable. The old operator overloading methods aren't deprecated. The new methods are more powerful, and that power is wasted in 100% of these cases.

[schveiguy]

This makes the code less readable.

What about with my suggested updates?

The old operator overloading methods aren't deprecated.

The old operator overloading methods are not documented (I will correct this), but should not be used in Phobos IMO. If there was ever a type that could use template + mixin opBinary, it would be Variant.

[DmitryOlshansky]

Seems useful if @schveiguy suggestions are followed through. @schuetzm any plans to update this?

[schuetzm]

This is ketmar's patch, I'm going to ask him. Maybe I'll also find some time to do it myself.

[JakobOvrum]

@schuetzm, any news on this?

[quickfur]

ping
Are we going ahead with this, or should we close it?

[schveiguy]

@quickfur I don't think it's worth merging this unless we take advantage of the templated nature of the new operators. No sense in doing this without doing it right. And I don't think ketmar will do it, so if @schuetzm isn't, I think we can just close it.