Iterator is needlessly a language item

[nagisa]

Iterator language item has been removed a year ago, but trait Iterator still has #[lang = "iterator"] on it. That’s not necessary and can be removed.

[apasel422]

Shouldn't both IntoIterator and Iterator be language items?

[GuillaumeGomez]

@apasel422: Seems more logical to remove it if it's not needed anymore than the contrary.

[apasel422]

My reasoning is that for depends on both traits (plus Option). Compiling this:

#![feature(lang_items)]
#![feature(no_core)]

#![crate_type = "lib"]

#![no_std]
#![no_core]

#[lang = "sized"]
trait Sized {}

pub fn foo() {
    for _ in () {
    }
}

yields

foo.rs:13:5: 13:9 error: failed to resolve. Maybe a missing `extern crate iter`? [E0433]
foo.rs:13     for _ in () {
              ^~~~
foo.rs:13:5: 13:9 help: run `rustc --explain E0433` to see a detailed explanation
foo.rs:13:5: 14:6 error: unresolved name `iter::IntoIterator::into_iter` [E0425]
foo.rs:13     for _ in () {
foo.rs:14     }
foo.rs:13:5: 14:6 help: run `rustc --explain E0425` to see a detailed explanation
foo.rs:13:5: 13:9 error: failed to resolve. Maybe a missing `extern crate iter`? [E0433]
foo.rs:13     for _ in () {
              ^~~~
foo.rs:13:5: 13:9 help: run `rustc --explain E0433` to see a detailed explanation
foo.rs:13:5: 14:6 error: unresolved name `iter::Iterator::next` [E0425]
foo.rs:13     for _ in () {
foo.rs:14     }
foo.rs:13:5: 14:6 help: run `rustc --explain E0425` to see a detailed explanation
foo.rs:13:5: 14:6 error: unresolved enum variant, struct or const `Some` [E0419]
foo.rs:13     for _ in () {
foo.rs:14     }
foo.rs:13:5: 14:6 help: run `rustc --explain E0419` to see a detailed explanation
foo.rs:13:5: 14:6 error: unresolved enum variant, struct or const `None` [E0419]
foo.rs:13     for _ in () {
foo.rs:14     }

[GuillaumeGomez]

Then why IntoIterator has been removed from language items?

[durka]

I think I agree with @apasel422... when for loops are desugared the absolute path to IntoIterator is used by name, but it seems like it would be more robust to use a lang item (cc #30809).

[nagisa]

Shouldn't both IntoIterator and Iterator be language items?

I generally would agree, but making them language items would be a breaking change of sorts and discussion towards that effect should be through a RFC. Removing the #[lang="iterator"] attribute now (i.e. cleanup of the libcore) and (re-)introducing Iterator and IntoIterator language items later (so the for-loop desugaring is more resilient) are not mutually exclusive goals and the former can easily land today, without waiting for the RFC or its outcome.

[apasel422]

@nagisa Proposed in rust-lang/rfcs#1457.

[alexcrichton]

I think it'd be fine to add these as language items to improve error messages, but in the meantime it seems fine to remove the annotation if they're not used in the compiler currently (e.g. we can add it back very easily at any point in the future)