Do not lint range_plus_one inside implementations of Index<RangeInclusive<_>>

[Darksonn]

When implementing indexing on RangeInclusive, we obviously cannot replace start .. end+1 with start ..= end, as that would be an infinite loop.

My situation is something like the following:

struct MyVec {
    inner: Vec<i64>,
}

impl Index<Range<usize>> for MyVec {
    type Output = [i64];
    fn index(&self, i: Range<usize>) -> &[i64] {
        &self.inner[i.start .. i.end]
    }
}

impl Index<RangeInclusive<usize>> for MyVec {
    type Output = [i64];
    fn index(&self, i: RangeInclusive<usize>) -> &[i64] {
        self.index(*i.start()..(*i.end() + 1))
    }
}

Here the impl of RangeInclusive delegates to the impl of Range by turning the inclusive range into an exclusive range.

Of course the same goes for IndexMut.

The warning is:

warning: an inclusive range would be more readable                                                                                                                                             
   --> src/lib.rs:15:20                                                                                                                                                                       
    |                                                                                                                                                                                          
 15 |         self.index(*i.start()..(*i.end() + 1))                                                                                                                                           
    |                    ^^^^^^^^^^^^^^^^^^^^^^^^^^ help: use: `*i.start()..=*i.end()`                                                                                                         
    |                                                                                                                                                                                          
    = note: #[warn(clippy::range_plus_one)] on by default                                                                                                                                      
    = help: for further information visit https://rust-lang-nursery.github.io/rust-clippy/master/index.html#range_plus_one 

Version:

$ cargo clippy -V
clippy 0.0.212 (2e26fdc 2018-11-22)

[detrumi]

Just checking whether ..= is used inside a .. impl isn't enough to cover all cases (though it might be good enough), since it could call other functions that use ..= instead. Would be hard to check against that, though.

[KamilaBorowska]

Maybe it would make sense to change the lint to detect such ranges being directly passing a function/method expecting an implementation of Iterator/IntoIterator/RangeBounds trait. If a function accepts one of those traits, then barring bugs, ranges should be identical to inclusive ranges. Such an implementation would also fix #3307.