unnecessary_fold: .any and .all may not always be safe replacements of .fold

[ghost]

Clippy version: 0.0.212
https://rust-lang-nursery.github.io/rust-clippy/v0.0.212/index.html#unnecessary_fold

Both .any and .all are short-circuiting while .fold evaluates every element in the iterator, if the iterator incurs side effects these two methods will have different behavior than .fold.

For example: https://play.rust-lang.org/?version=stable&mode=debug&edition=2015&gist=893a485dc791c89d95d6ce6f744af9bf

[camsteffen]

I just ran into this.

I think the short-circuiting methods (any and all) should be removed from this lint. If they are still wanted, they should be moved to a pedantic lint with known false positives.

At a minimum, the warning message should have a note like "this may change semantics" or "this may change side-effects".

[camsteffen]

I suppose it may be possible to detect cases where there are side-effects?

[maartendeprez]

Hi, I also ran into this. Clippy suggests replacing .fold(false, |a,b| a || b) with .any(...), but applying the suggested fix would lead to subtle bugs in my code because of side effects in the iterator. I think this lint (in its current version) should at least be MaybeIncorrect and explain the differences.

[falsetru]

I also ran into this.

https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=c74eadf27c94c189d1649b7f7a4c187c

[dbeckwith]

Also just noticed this. I think it's good to mention any as a possible alternative to fold but it should be very clear about the difference in short-circuiting behavior.