Put impls of primitives near the traits they implement in std

[brson]

In std there are modules for the primitive types and these modules tend to contain the impls for several of the most 'core' Rust traits. I would rather these live in the module that declares the trait. std is going to continue to be refactored, and the crates these traits live in may change. Having the primitive impls in one place makes that refactoring easier.

[brson]

This also lets us delete some modules that are nothing but impls.

[Kimundi]

I started working on this

[Kimundi]

Turns out I read the description wrong and spend the whole day refactoring primitive impls away from their traits.

@brson , I don't suppose this could still be useful: https://github.com/Kimundi/rust/compare/move_impls?expand=1 ?

[brson]

@Kimundi That is indeed the opposite of what I wanted :(

[malu]

If this is still relevant (it seems to be) and @Kimundi stopped working on this I could work on this.

[Kimundi]

@malu Sorry, should've made it clear that I'm not working on it any more.

[kud1ing]

I've always found it surprising that implementations for independent types are bundled in the module where the interface is defined.

Is refactorability the only reason for this?

[thestinger]

It's very convenient because you can usually reuse a lot of code with a macro. It's not relevant to the user-facing API at all.

[sfackler]

You also avoid ending up with blank modules like this: http://static.rust-lang.org/doc/master/std/unit/index.html

[malu]

@kud1ing I guess it's also nice to have impls for standard types near trait def because it like saying "Look, this trait is actually useful and can be fulfilled easily by these standard types."

[tbu-]

@malu Still working on it?

[malu]

@tbu- Yes. Sorry, the past day were pretty busy, I hope I can do a pull request this weekend after rebasing.

[malu]

Okay, something went horribly wrong during rebasing. I don't think I'm going to be able to fix this anytime soon. :(
So anybody should feel free to work on this.

[flaper87]

@malu sorry to hear that :( Any chance you can upload your work pre-rebase so that people working on this change can take over based on your changes (or at least use them as reference) ?

[huonw]

You can use git reflog to "reset" to a previous branch state, restoring you back to how it was before.

[ahmedcharles]

Is this still worth working on?

[Sawyer47]

I could work on this. I think it's worth it: core::bool would go away (to_bit function is pretty useless), core::num::{i*,u*,f*} would only contain constants and could be easily removed in the future. Plus implementing one trait for many types with macro will probably result in shorter code.

(well, I started working on this, let me know if someone else is working on this too)