New lint: iter_overeager_cloned

[pmnoxx]

What it does

Calling .iter().cloned.last() involved doing unnecessary work.
It's better to write it as .iter().last().cloned()

Same thing can rule can be applied to:

• .iter().cloned().next()
• .iter().cloned().count()
• .iter().cloned().flatten()
• .iter().cloned().take(...)
• .iter().cloned().skip(...)
• .iter().cloned().nth(...)

Lint Name

cloned_last

Category

style, perf

Advantage

• More efficient code

Drawbacks

• I don't see any

Example

    let _: Option<String> = vec!["1".to_string(), "2".to_string(), "3".to_string()]
        .iter().cloned().last()

Could be written as:

    let _: Option<String> = vec!["1".to_string(), "2".to_string(), "3".to_string()]
        .iter().last().cloned()

[pmnoxx]

I created a new PR for that lint: #8203

[kornelski]

Iterators are lazy, so .iter().cloned().take(...) and .iter().cloned().next() are perfectly fine.

[llogiq]

By themselves .cloned().take() is ok, but that could be followed by .last() and that would trigger the lint next time.

With .next() I somewhat agree, though I note that it will still produce more code to optimize away, so it will likely at least positively affect compile time (and probably debug runtime). Is there any benefit to not switching the order?

[kornelski]

Ok, I see your point. The benefit would be only in number of warnings from clippy for a not-strictly-necessary case.

[robertbastian]

I think the examples for this are bad and should be updated:

// Bad
vec.iter().cloned().take(10);

// Good
vec.iter().take(10).cloned();

These are equivalent, not a single clone happens. Even if we add .next(), only one clone will happen in both cases. collect would show it, or using skip:

// Bad
vec.iter().cloned().skip(10).next(); // 11 clones

// Good
vec.iter().skip(10).cloned().next(); // 0-1 clones

---

// Bad
vec.iter().cloned().last();

// Good
vec.iter().last().cloned();

This has a typo, the good line should be .clone().

[llogiq]

No. Option::cloned() is the correct function to call here, and that's what the suggested code does.

[robertbastian]

What about the first part of my comment?