Interface narrowings don't yet work

[jyasskin]

Currently:

type base =
  obj {
    fn foo();
  };
obj derived() {
  fn foo() {}
  fn bar() {}
}
fn main() {
  let derived d = derived();
  let base b = d;
}

fails to compile with:

test2.rs:11:19 - 11:20:E:Error: mismatched types: expected obj {
    fn foo() -> ();
    fn bar() -> ();
} but found obj {
    fn foo() -> ();
}

Adding an explicit cast to base crashes the compiler. The rust docs say "Such a “plain” object type can be used to describe an interface that a variety of particular objects may conform to, by supporting a superset of the methods." so I'd expect the implicit conversion to work. (On the other hand, then "any" is just "obj {}", and Graydon has said he doesn't want an implicit conversion to "any".)

[graydon]

The mechanism I have sketched but only partly implemented for this would be 'let base b = d as base;', but it doesn't work yet. Most of the machinery is there, it's just incomplete.

When you say "an explicit cast crashes the compiler", what notation were you using for this?

[jyasskin]

I was trying
let base b = base(d);
by analogy with uint(3).

[graydon]

Shifted to rustc.

[lkuper]

What's the status of this issue with regard to the current state of rustc and the object system?

[lkuper]

Just pushed an xfail'd test for this issue. The error we're getting now is:

../rust/src/test/run-pass/interface-narrowing.rs:18:15:18:24: error: non-scalar cast: <T1> as obj {
    fn foo();
}
../rust/src/test/run-pass/interface-narrowing.rs:18   let b:base = d as base;
                                                                   ^~~~~~~~~

I'm un-assigning myself from this one and marking it as unassigned since, realistically speaking, I don't think I'll have time to handle it in the next four days. It's up for grabs...

[marijnh]

Objs are gone. Closing.