Type inference of integer literals: inconsistency with the reference

[jaystrictor]

The reference currently says the following for integer literals

The type of an unsuffixed integer literal is determined by type inference:
[...]
If the program context under-constrains the type, it defaults to the signed 32-bit integer i32.

This does however not work for some inherent methods of signed integer types:

trait A {
    fn what_type(&self);
}


impl A for i16 {
    fn what_type(&self) {
        println!("i16");
    }
}

impl A for i32 {
    fn what_type(&self)  {
        println!("i32");
    }
}

fn main() {
    let z = 1;
    z.what_type();
    //z.is_positive(); // <- uncomment this line
}

As you can see, z is under-constrained in this case. The compiler does the right thing and defaults to type i32, so calling the trait method what_type() works.

If you uncomment the inherent method call to is_positive(), z still is under-constrained in pretty much the same way as before, however, the compiler fails to default to i32 and instead prints

error: no method named `is_positive` found for type `{integer}` in the current scope

---

Also note that the reference says

If an integer type can be uniquely determined from the surrounding program context, the unsuffixed integer literal has that type.

But the following example show that the order of statements is important:

fn one() { // compiles
    let a = 0;
    take_some(a);
    a.is_positive();
}

fn two() { // does not compile
    let a = 0;
    a.is_positive();
    take_some(a);
}

fn take_some(var: i32) {
    // do nothing
}

---

I think floating-point literals have the same issue, but I haven't checked.

Issue #39255 and issue #40985 may be related.

[steveklabnik]

i've moved this over to rust-lang/reference#104