new lint: Recommend using slice.into() to box the slice

[Robbepop]

What it does

Given a slice like

let x: &[i32] = &[1, 2, 3, 4, ..];

When users write something like

let y: Box<[i32]> = x.iter().copied().collect();

Then clippy should propose to replace x.iter().copied().collect() with x.into(). This is not only shorter and more readable but also more efficient according to my benchmarks since x.iter() can make use of memset. So users should write

let y: Box<[i32]> = x.into();

instead.

I often come across this anti-pattern of boxing up a slice in codebases.

I don't know if this explodes the scope of this issue but technically we could also lint against the following usage with similar reasoning: When users write something like

let y: Box<[i32]> = x.to_vec().into_boxed_slice();

Then clippy again should propose to replace x.to_vec().into_boxed_slice() with x.into() since x.to_vec().into_boxed_slice() creates a Vec with some capacity that fits x but the call into_boxed_slice does not guarantee that the former won't be shrunk which causes another memory reallocation. Whereas using x.into() won't have this problem.

Lint Name

slice_into_boxed_slice

Category

perf

Advantage

• The replacement with x.into() is less verbose and likely more readable than the alternatives.
• In the first example x.into() can make use of memset under the hood which yields better performance according to my benchmarks.
• In the second example x.into() will guarantee to only cause at most one memory allocation whereas x.to_vec().into_boxed_slice() might perform more than that.

Drawbacks

None?

Example

Given a slice like

let x: &[i32] = &[1, 2, 3, 4, ..];

Then

let y: Box<[i32]> = x.iter().copied().collect();

And

let y: Box<[i32]> = x.to_vec().into_boxed_slice();

Could both be written as:

let y: Box<[i32]> = x.into();

[HMPerson1]

@rustbot claim