operator overloading outside the type

[Walter Bright]

On 3/29/2025 9:25 AM, Timon Gehr wrote:
> Yes. UFCS never overrides a matching member function, no matter how well the 
> UFCS function matches. UFCS is not even attempted if a call can be resolved via 
> member function.

That's correct. The issue brought up in my preceding post is *should* operators 
be treated as if they were functions? Or are they special? I've seen what 
happens when they are treated as functions. People write code that looks like 
conventional C++ code, but with the operators overloaded to do something utterly 
different. They'd write articles saying how cool this was. I had the opposite 
reaction.

There are a number of deliberate characteristics in D that discourage writing 
incomprehensible code, like no macros and no version algebra. The operator 
overloading capability in D is deliberately restrictive, to make it harder to do 
non-arithmetic things with it.

With functions, there are no implicit expectation of what is does, we all know 
that. But the expectation with operators is `a+b` does an addition. (That's why 
D uses ~ for concatenation, not +) If the user suspects that it does something 
else, at least he can visit the type declaration to see if it was repurposed.

But overloading operators with freestanding functions? It becomes a nightmare 
with non-trivial programs.