Rust Analyzer passive or doesn't locate available methods and macros

[wmstack]

rust-analyzer version: rust-analyzer version: 0.3.2291-standalone (f3998f7f8a 2025-02-02)

rustc version: rustc 1.84.1 (e71f9a9a9 2025-01-27)

editor or extension: Vscode Rust Analyzer rust-lang.rust-analyzer

relevant settings: Regular Programming in a Simple Crate with cargo init

code snippet to reproduce:

use core::str;


fn main() {
  str::fr
}

Expected output

Actual Output

Issue

Rust Analyzer refuses to look into core::str for methods even though this would be acceptable by the Rust compiler to use directly.

Related Issue

use core::str;
use std::{
  io::stdout,
  process::Command,
};

fn main() {
  let output = Command::new("ls")
    .arg("-a")
    .arg("--color")
    .output()
    .unwrap();
    
    println!("{}", str::from_utf8(output.stdout.refm^));
}

Macros ref and refm are not visible when writing .ref. It is kind of impossible to access .ref

[wmstack]

Related Discussion: https://users.rust-lang.org/t/rust-analyzer-works-most-of-the-time/81114

[ChayimFriedman2]

This is not as easy. For one, str can resolve to either core::str or the builtin type str - and while we do have the ability in r-a to make it resolve to either (and in fact we resolve it correctly in analysis), here we will need to resolve it as both, and that (I imagine) will require some deep changes (although perhaps we can have some completion-specific hack).

See also #19088, although it is about resolving it after it's fully typed, which is easier.

[wmstack]

Okay I am less concerned by str constructors as it seems they are going to be merged with the primitive type:
rust-lang/rust#131114

However, I am more concerned with why the .ref macro doesn't consistently show up, in a field where usually Rust Analyzer knows a reference is suitable?

Desirable Behaviour (.ref not even needed as &stdout field shows)

Undesirable Behaviour 1: Rust Analyzer forgets &stdout is useful

Follow Up Attempt to invoke .ref macro:

The only difference is first one is direct, and the other is in a macro println! call.

[ChayimFriedman2]

The problem with ref is that output.stdout.ref is not a valid expression (due to ref being a keyword), therefore macro expansion fails, therefore we cannot provide completions. It does complete (down the list) when you type re (or refm, as you noticed yourself). This is something we can't fix easily, but we expect to fix once the switch to new Salsa is done and its DB forking feature is implemented.

The problem with & not auto-suggested inside a macro is a different issue - can you open a new issue for it?