merge keywords: impl and implements

[graydon]

I would prefer to see class C impl D rather than class C implements D. The latter feels redundant and verbose, and I think the former is unambiguous.

[catamorphism]

I'm fine with changing "implements" to something else, but using impl there just makes me shudder. I don't mind short keywords, but using one together with a non-abbreviated keyword is like fingernails on a blackboard to me. (I acknowledge this comment as a total bikeshed.)

[brson]

Classes and impls should use the same keyword to say what iface they are implementing. impls currently use of, but class duck of quackable doesn't sound that great.

[lkuper]

How about class C impls D? It's short, but still clearly "implements" rather than "implementation".

[catamorphism]

@lkuper : +1. @brson : I agree in principle, but can't think of a good syntax either. We might just have to bite the bullet and continue to have an additional keyword. I prefer readability to consistency, but that's just my opinion.

[graydon]

I think niko suggested : on irc?

[catamorphism]

Oh yeah, now I remember that. Sure, ':' is ok. I haven't seen any objections to it, so will implement that.

[lkuper]

I like : as well.

[graydon]

@brson and @nikomatsakis both indicated on IRC support for this, with the revised syntax:

impl for ty: iface { ... }
impl foo for ty: iface { ... }
class foo: iface { ... }

[catamorphism]

I like it. Will do.

[kud1ing]

Two downsides, which do not really matter:

• "class X : Y" will make some people compare Rust to C++
• ":" has worse grep-ability than "impls" or "implements"

[catamorphism]

@Lenny222 I don't care about the first one that much; I do think the second one matters, but we should probably have better IDE-ish tools anyway that would obviate the need for grepping :-)

[graydon]

The headline version of this is fixed and the rest (use of colon and such) is something @lkuper is going to pick up during #2794, so closing.