Allow using Self in impls

[emberian]

This is convenient because it allows you to copy-paste trait declarations.

[emberian]

Fully back-compat, certainly not a 1.0 priority.

[emberian]

This would essentially involve adding a "hidden" type parameter in every impl.

[emberian]

cc @nikomatsakis @pnkfelix

[eddyb]

Is there really a need for another magic type parameter?
I would think astconv can resolve Self in impls to the actual type, and nothing needs to see ty_self.

[emberian]

Sure that could work

On Thu, Feb 27, 2014 at 3:30 PM, Eduard Burtescu
[email protected]:

Is there really a need for another magic type parameter?
I would think astconv can resolve Self in impls to the actual type, and
nothing needs to see ty_self.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/12600#issuecomment-36287579
.

[nikomatsakis]

As @eddyb said, I think this is basically syntactic sugar, and I have no particular problem with it. I think the main reason not to do it (and the reason we didn't before) is so as not to obscure the role of Self as a type parameter and so on, which isn't applicable to impls, but I don't know that it really matters.

[pnkfelix]

+1 to all of what Niko said. In particular, I do like the sugar being suggested here. But at the same time, I noticed that in a recent rust-dev post, @anasazi used the notation Self<T> to apply a hypothetical higher-kinded Self : * -> * to a type parameter. If we wanted to leave such options open for the future, it may constrains how we use/abuse Self now. (But as niko said, the role of Self in trait definitions may not be relevant to its role in impls.)

[flaper87]

This should probably go through the new RFC process. @cmr are you going to submit one? Can we close this issue?

[emberian]

I suppose I can write one later this week.

[nikomatsakis]

On Mon, Mar 24, 2014 at 04:10:00PM -0700, Flavio Percoco Premoli wrote:

This should probably go through the new RFC process. @cmr are you going to submit an one? Can we close this issue?

Agreed

[rust-highfive]

This issue has been moved to the RFCs repo: rust-lang/rfcs#310