Eliminate $crate in --pretty=expanded

[dtolnay]

Similar to #37637 so cc @jseyfried.

cat src/main.rs

fn main() {
    println!("rust");
}

rustc --pretty=expanded -Zunstable-options src/main.rs

#![feature(prelude_import)]
#![no_std]
#[prelude_import]
use std::prelude::v1::*;
#[macro_use]
extern crate std as std;
fn main() {
    $crate::io::_print(::std::fmt::Arguments::new_v1({
                                                         static __STATIC_FMTSTR:
                                                                &'static [&'static str]
                                                                =
                                                             &["rust\n"];
                                                         __STATIC_FMTSTR
                                                     },
                                                     &match () {
                                                          () => [],
                                                      }));

The $crate prevents the expanded code from being parsed by rustfmt.

rustc --pretty=expanded -Zunstable-options src/main.rs | rustfmt

error: expected expression, found `$`
 --> stdin:8:5
  |
8 |     $crate::io::_print(::std::fmt::Arguments::new_v1({
  |     ^

This breaks cargo expand which runs the code through rustfmt by default.

I think --pretty=expanded should produce valid Rust code which means either eliminate $crate like we did for #37637, or allow rustfmt to parse it.

[nagisa]

Why are you relying on unstable features for, umm, stable tooling? --pretty et al by now are only used for debugging purposes and won’t be stabilised.

[bluss]

I don't think expand can produce valid Rust code (how does it handle hygiene?)

[dtolnay]

Before the $crate change, it produced syntactically valid Rust code even if the semantics were different in the case of hygiene collisions. This was useful for debugging purposes as @nagisa points out.

Since the $crate change, it produces syntactically invalid code which I consider a bug.

[jseyfried]

Fixed in #38232 (#38232 (comment)).
After that lands, we will pretty-print $crate::io::_print as ::io::_print.

[xiaoniu-578fa6bff964d005]

$crate shows up again!

rustc +nightly-2018-06-13-x86_64-unknown-linux-gnu --pretty=expanded -Zunstable-options main.rs

#![feature(prelude_import)]
#![no_std]
#[prelude_import]
use std::prelude::v1::*;
#[macro_use]
extern crate std;
fn main() {
    ::io::_print(::std::fmt::Arguments::new_v1(&["rust\n"],
                                               &match () { () => [], }));
}

rustc +nightly --pretty=expanded -Zunstable-options main.rs

#![feature(prelude_import)]
#![no_std]
#[prelude_import]
use ::std::prelude::v1::*;
#[macro_use]
extern crate std;
fn main() {
    {
        $crate::io::_print($crate::fmt::Arguments::new_v1(&["rust\n"],
                                                          &match () {
                                                               () => [],
                                                           }));
    };
}

rustc +nightly --version

rustc 1.33.0-nightly (0c0c58528 2019-01-19)

[petrochenkov]

Caused by #57155.
Reopening and assigning to myself.

[petrochenkov]

Fixed in #57915