New lint: Bound by BufRead instad of bounding by Read and then instantiating BufReader

What it does

If a generic function takes R: io::Read and then wraps it in io::BufReader warn that this may introduce unnecessary allocation and copying. Recommend io::BufRead bound instead.

Categories (optional)

• Kind: perf

If the caller already has a type implementing BufRead then wrapping introduces allocation and also additional copying of data between the buffers.

Drawbacks

The change is backwards-incompatible and major version must be bumped if the bound is in public API. All callers that do not have buffered readers need to wrap the reader in BufReader themselves.

Example

fn parse_foo<R: io::Read>(reader: R) -> io::Result<Foo> {
    let mut reader = io::BufReader::new(reader);

    // Rest of the parsing logic ...
}

Could be written as:

fn parse_foo<R: io::BufRead>(mut reader: R) -> io::Result<Foo> {

    // Rest of the parsing logic ...
}

The issue provides Rust examples but names no implementation files, tests, or entry point. Start by locating the existing lint implementation and its tests in rust-clippy, then verify the lint detects generic io::Read parameters wrapped in io::BufReader and recommends an io::BufRead bound, including the stated public-API compatibility concern.