Rust: Ownership problem when trying to get next byte from Stdin - rust

Question

I was modifing a code snippet from github which enable fast i/o in Rust for competitive-programming.
I want to add a feature making it can read single character from stdin, just like getchar in C. My approach was a combination of the pub fn read<T: std::str::FromStr> in the original code snippet and this SO answer
//////////////////////////////////////////////////////////////////////
/// https://github.com/kenkoooo/competitive-programming-rs/blob/master/src/utils/scanner.rs
/// let (stdin, stdout) = (std::io::stdin(), std::io::stdout());
/// let mut sc = IO::new(stdin.lock(), stdout.lock());
pub struct IO<R, W: std::io::Write>(R, std::io::BufWriter<W>);
impl<R: std::io::Read, W: std::io::Write> IO<R, W> {
pub fn new(r: R, w: W) -> IO<R, W> {
IO(r, std::io::BufWriter::new(w))
}
pub fn write<S: ToString>(&mut self, s: S) {
use std::io::Write;
self.1.write_all(s.to_string().as_bytes()).unwrap();
}
pub fn read<T: std::str::FromStr>(&mut self) -> T {
use std::io::Read;
let buf = self
.0
.by_ref()
.bytes()
.map(|b| b.unwrap())
.skip_while(|&b| b == b' ' || b == b'\n' || b == b'\r' || b == b'\t')
.take_while(|&b| b != b' ' && b != b'\n' && b != b'\r' && b != b'\t')
.collect::<Vec<_>>();
unsafe { std::str::from_utf8_unchecked(&buf) }
.parse()
.ok()
.expect("Parse error.")
}
pub fn usize0(&mut self) -> usize {
self.read::<usize>() - 1
}
pub fn vec<T: std::str::FromStr>(&mut self, n: usize) -> Vec<T> {
(0..n).map(|_| self.read()).collect()
}
pub fn chars(&mut self) -> Vec<char> {
self.read::<String>().chars().collect()
}
pub fn char(&mut self) -> char {
self
.0
.by_ref()
.bytes()
.next()
.unwrap()
.unwrap() as char
}
}
///////////////////////////////////////////////////////////////////////
#[allow(non_snake_case)]
fn main() {
let (stdin, stdout) = (std::io::stdin(), std::io::stdout());
let mut sc = IO::new(stdin.lock(), stdout.lock());
let c = sc.char();
sc.write(c);
}
cargo run and the output was:
$ cargo run
warning: unused manifest key: package.author
Compiling ralgo v0.1.0 (/home/xxx/ralgo)
error[E0507]: cannot move out of a mutable reference
--> src/main.rs:40:9
|
40 | / self
41 | | .0
42 | | .by_ref()
43 | | .bytes()
| | ------^
| |__________|_____|
| | move occurs because value has type `R`, which does not implement the `Copy` trait
| value moved due to this method call
|
note: `bytes` takes ownership of the receiver `self`, which moves value
--> /home/xxx/.rustup/toolchains/nightly-x86_64-unknown-linux-gnu/lib/rustlib/src/rust/library/std/src/io/mod.rs:922:14
|
922 | fn bytes(self) -> Bytes<Self>
| ^^^^
For more information about this error, try `rustc --explain E0507`.
error: could not compile `ralgo` due to previous error
My questions:
The reason for this error and how to fix it.
Is there any better way to achieve getchar in Rust?
Why the rust-analyser shows the return type of let buf = self.0.by_ref().bytes() in the original read function is Bytes<&mut R> but for my code self.0.by_ref().bytes() it was Bytes<R>?

Answer 1

Upfront, I want to note that your read is unsound. And furthermore the cost of allocating a vector is almost certainly going to dwarf the cost of validating that said vector is UTF8, so it's rather unnecessary.
The reason for this error and how to fix it.
Because you don't use std::io::Read globally, the compiler knows that R specifically implements Read (as that's a bound), it does not know about any other implementation of Read. Notably, it does not know about impl<R: Read + ?Sized> Read for &mut R. Thus as far as it is concerned, the closest thing which would have a bytes method is the original reader, which it tries to deref'. The error is much clearer if you assign the result of by_ref to a local variable.
Just stop using fully qualified path like this, especially for traits. Your code is strictly less readable than if you just imported Read, Write, and BufWriter at the top.
Is there any better way to achieve getchar in Rust?
getchar is a horrible function so I'm not too clear on why you'd want that. But I would suggest just using byteorder's read_u8.
Why the rust-analyser shows the return type
Possibly because it's getting confused and assumes all the traits are in scope.
.skip_while(|&b| b == b' ' || b == b'\n' || b == b'\r' || b == b'\t')
.take_while(|&b| b != b' ' && b != b'\n' && b != b'\r' && b != b'\t')
These seem like complicated ways of not calling u8::is_ascii_whitespace.