Make sure type inference with a..b as good as range(a,b)

[edwardw]

The new ::ops::Range has separated implementations for each of the
numeric types, while the old ::iter::Range has one for type Int.
However, we do not take output bindings into account when selecting
traits. So it confuses typeck and makes the new range does not work as
good as the old one when it comes to type inference.

This patch implements Iterator for the new range for one type Int.
This limitation could be lifted, however, if we ever reconsider the
output types' role in type inference.

Closes #21595
Closes #21649
Closes #21672

[rust-highfive]

r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

[ghost]

I am curious what happens with a negative range eg. -1..10
I'll test this with your PR and let you know. Currently started compiling rust locally.

[edwardw]

It should work as expected since after type inference, the numeric fallback will kick in and -1..10 will then be equivalent to -1i32..10.

[bluss]

Does this still allow impl Iterator for Range<MyLocalType> ? I don't think that's something we require or designed for, but I'm curious nonetheless.

[alexcrichton]

Unfortunately with the definition as-is I don't think that we can have this ExactSizeIterator impl here. As defined it's possible for size_hint to return (0, None) which goes against the contract of ExactSizeIterator.

While we don't have negative bounds, perhaps the ExactSizeIterator trait could still just be manually defined for each primitive type?

[edwardw]

Done, the hybrid implementations for range work!

[edwardw]

Hmm, good question. I think so offhand, as long as Range<T: Int> and Range<MyLocalType> are not considered conflict implementations.

[ghost]

my local compilation reports this:

test [compile-fail] compile-fail/xc-private-method2.rs ... ok
test [compile-fail] compile-fail/xcrate-private-by-default.rs ... ok

failures:

---- [compile-fail] compile-fail/range-1.rs stdout ----

error: expected error on line 19 not found: `core::iter::Iterator` is not implemented for the type `core::ops::Range<f32>`
status: exit code: 101
command: x86_64-unknown-linux-gnu/stage2/bin/rustc /home/emacs/build/rust/src/test/compile-fail/range-1.rs -L x86_64-unknown-linux-gnu/test/compile-fail --target=x86_64-unknown-linux-gnu -L x86_64-unknown-linux-gnu/test/compile-fail/range-1.stage2-x86_64-unknown-linux-gnulibaux -C prefer-dynamic -o x86_64-unknown-linux-gnu/test/compile-fail/range-1.stage2-x86_64-unknown-linux-gnu --cfg rtopt -O -L x86_64-unknown-linux-gnu/rt
stdout:
------------------------------------------

------------------------------------------
stderr:
------------------------------------------
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:15:13: 15:23 error: start and end of range have incompatible types: expected `usize`, found `isize` (expected usize, found isize) [E0308]
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:15     let _ = 0us..10is;
                                                                       ^~~~~~~~~~
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:19:5: 26:8 error: the trait `core::num::Int` is not implemented for the type `f32` [E0277]
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:19     for i in 0f32..42f32 {}
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:20     //~^ ERROR `core::iter::Iterator` is not implemented for the type `core::ops::Range<f32>`
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:21     //~^^ ERROR
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:22     //~^^^ ERROR
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:23     // FIXME(#21528) not fulfilled obligation error should be reported once, not thrice
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:24 
                                                           ...
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:27:17: 27:24 error: the trait `core::marker::Sized` is not implemented for the type `[usize]` [E0277]
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:27     let range = *arr..;
                                                                           ^~~~~~~
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:27:17: 27:24 note: `[usize]` does not have a constant size known at compile-time
/home/emacs/build/rust/src/test/compile-fail/range-1.rs:27     let range = *arr..;
                                                                           ^~~~~~~
error: aborting due to 3 previous errors

------------------------------------------

thread '[compile-fail] compile-fail/range-1.rs' panicked at 'explicit panic', /home/emacs/build/rust/src/compiletest/runtest.rs:1458



failures:
    [compile-fail] compile-fail/range-1.rs

test result: FAILED. 1643 passed; 1 failed; 22 ignored; 0 measured

thread '<main>' panicked at 'Some tests failed', /home/emacs/build/rust/src/compiletest/compiletest.rs:252
/home/emacs/build/rust/mk/tests.mk:729: recipe for target 'tmp/check-stage2-T-x86_64-unknown-linux-gnu-H-x86_64-unknown-linux-gnu-cfail.ok' failed
make: *** [tmp/check-stage2-T-x86_64-unknown-linux-gnu-H-x86_64-unknown-linux-gnu-cfail.ok] Error 101

[edwardw]

Thanks @emanueLczirai, the fallout has been fixed.

[edwardw]

r? @alexcrichton

[ghost]

Compiles successfully for me now. Thanks! (and numeric fallback is as you said earlier: i32)

One question: is that commented FIXME that you removed no longer necessary there? as in, is it fixed or no longer makes sense due to this PR?

[edwardw]

It is simply not there any more. I doubt it has anything to do with this PR though.

[alexcrichton]

@bors: r+ b9c055c

Thanks!

[bors]

⌛ Testing commit b9c055c with merge c2bda2a...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-32-nopt-t, auto-win-32-opt, auto-win-64-nopt-t, auto-win-64-opt