New lint: slice_indexing_after_deconstruction

[xFrednet]

What it does

The lint checks for slices that come from value deconstruction and are only used to access individual slice values.

Categories (optional)

• Kind: pedantic (This could also be style, but I'm not sure if we want it warn-by-default.)

Advantages

Deconstruction of slices allows direct value access without using indexes on the slice, which can lead to panics. It also enables the user to give slice values a better name and indirectly check the correct length of the slice. See #7566 files as an example of how this can be an improvement.

Clippy already had a few instances where ICEs would have been prevented by using deconstruction of slices.

Example

fn main() {
    let slice: Option<&[u32]> = Some(&[1, 2, 3]);
    
    if let Some(slice) = slice {
        println!("{}", slice[0]);
    }
}

Could be written as:

fn main() {
    let slice: Option<&[u32]> = Some(&[1, 2, 3]);
    
    if let Some([first, ..]) = slice {
        println!("{}", first);
    }
}

Notes:

• Clippy uses pattern matching extensively, often in if_chain macros with if let statements. I think it would be reasonable for this lint to also check if statements inside if_chain macros.
• The lint probably shouldn't warn if the slice instance is used for something else besides direct value access.

[camsteffen]

We have to pick some max threshold for indexes to lint. Like we don't want to lint slice[100]. Maybe slice[3] is a reasonable max to lint? In other words, suggest a slice pattern with up to 4 elements.

[xFrednet]

Good suggestion, a threshold of 3 would catch all instance in Clippy that I've seen. We could also have this configurable and have the default be 3 👍.

[xFrednet]

Reviewing a rustup PR I found this awesome refactoring which could also be suggested by this lint as an enhancement later on:

// from
for win in blocks.windows(2) {

// to
let mut iter = blocks.windows(2);
while let Some(&[win0, win1]) = iter.next() {

(Let's just ignore that this refactoring was done in my first bigger lint implementation xD)

[camsteffen]

I'm holding out for for [a, b] in slice.array_windows(2) (currently unstable)

[xFrednet]

That would be even nicer. I want to work on this once rustup is merged. It sounds like a useful and fun lint 🙃

@rustbot claim