Factual error in PartialEq documentation

[QuineDot]

The documentation for PartialEq states:

x.eq(y) can also be written x == y, and x.ne(y) can be written x != y.

This is not true, as #44762 illustrates, as well as this URLO-inspired example:

    let val = 1i32;
    let val_ref = &val;
    let is_equal = val.eq(val_ref); // Compiles
    let is_equal = val == val_ref; // Does not compile    

The other operator's documentation don't have this error. Shl (and Shr) also explicitly note some differences in inference between operators and the method call.

@rustbot label +A-docs

[camelid]

I assume this is because auto(de)ref kicks in for the method version but not the operator version.

[QuineDot]

I assume this is because auto(de)ref kicks in for the method version but not the operator version.

Yes (and experimenting with changing that is what #44762 is about).

[JanBeh]

The reference on comparison operators clarifies:

Unlike the arithmetic and logical operators above, these operators [the comparison operators] implicitly take shared borrows of their operands, evaluating them in place expression context:

# let a = 1;
# let b = 1;
a == b;
// is equivalent to
::std::cmp::PartialEq::eq(&a, &b);

This means that the operands don't have to be moved out of.

[QuineDot]

But that's also counterfactual!

// Commented code does not compile
fn main() {
    let v = 1;
    let x: *const u32 = &v;
    let y: &u32 = &1;
    let _ = dbg!(x == y);
    // let _ = dbg!(PartialEq::eq(x, y));
    // let _ = dbg!(PartialEq::eq(x, &y));
    // let _ = dbg!(PartialEq::eq(&x, y));
    // let _ = dbg!(PartialEq::eq(&x, &y));
    let _ = dbg!(PartialEq::eq(&x, &(y as _)));

    // But that can't be a desugaring...
    // let _ = dbg!(v == y);
    let _ = dbg!(&v == y);
    // let _ = dbg!(PartialEq::eq(&v, &(y as _)));
}

Operator resolution is its own thing with many niche differences and, in my opinion, all phrasing like "equivalent to" / "can be written as" / "is the same as" / "desugars" should be replaced with something more like "is notionally" / "is like" / etc, perhaps with a list of exceptions you're likely to hit.

Another example from the reference is compound assignment (+=):

• They mention a place expression requirement -- great!
• They mention order evaluation differs for primitives(!) -- great!
• But then they blunder and say

  the following expression statements in example are equivalent:

Even modulo the exceptions they did cover, the remaining cases are not equivalent because += can participate in two-phase borrowing.

// commented code does not compile
use core::num::Wrapping;
fn main() {
    let n = &mut [Wrapping(0)][..];
    n[0] += n[0];
    // std::ops::AddAssign::add_assign(&mut n[0], n[0]);
}

(See also.)

[hkBst]

@QuineDot it seems to me that you have accumulated enough knowledge around this issue to create a good PR for this. The documentation lives here: https://github.com/rust-lang/rust/blob/master/library/core/src/cmp.rs.