Deref not applying to things that implement Index

[steveklabnik]

This should compile, but doesn't:

use std::rc::Rc;

fn main() {
    let vec = Rc::new(vec![1i, 2, 3]);
    println!("{}", vec[0]);
}

[bluss]

Same with box instead of Rc.

[brson]

Nominated. @pcwalton What's the intended behavior here?

[nrc]

cc me

[nrc]

I would expect this to work. When we check to see if we can index we should search for the Index trait the same way we do when explicitly writing e.index(...), which would include applying implicit derefs, including custom ones.

It is backwards compatible though, so shouldn't need to block 1.0.

[pcwalton]

I think this is not a 1.0 issue.

[steveklabnik]

What's the reason this was removed? It's true that it's not backwards compatible, but it leads to really unfortunate code.

[aturon]

@steveklabnik The nomination hasn't been removed -- we'll discuss this at tomorrow's triage meeting to determine whether to put it on the 1.0 milestone. At that point we'll remove the nomination tag and replace it with priorities/milestones.

FWIW, I agree that this is really unfortunate as it stands. But the question for the milestone is: would we hold back the 1.0 release until it was fixed? The answer can be "no" even for very high priority items.

I do think that @nikomatsakis's work on method resolution, as part of trait reform, is likely to fix this well before 1.0.

[steveklabnik]

Oh yeah, and honestly, I'm willing to make sacrifices for a 1.0, I guess
what I mean is "I don't know how much work this actually is to fix."

[pnkfelix]

P-high, not a 1.0 blocker.

[nikomatsakis]

I think this'll probably wind up being fixed as fallout from various changes.