You could use the .scan() over iter to hold the state

a.iter().scan(init, |state, &x| {
    *state = f(*state, x);
    Some(*state)
}).collect()

Almost did the same thing...but I changed the trait bounds on array_diff to also require that T implement the Copy trait.
Then I used iter and copied instead of just using into_iter like I should have...because I initially misspelled into_iter while trying this exact thing :P

Yes, but array sizes are quite small in test cases, so creating additional structures can be less productive in this specific case.

This code has O(n*m) runtime as for each element we have to re-scan in b from scratch.

I never knew there was a min function on iter, so short and clever

I forgot about iterators on the array. I need to stop thinking of Rust arrays as C style arrays. I converted my array to a vec first but essentially my solution was the same.

nvm it was just a typo.

This compiles and runs perfectly. Could you copy your code here?

i tried this and kept getting an error when a[] was empty

I like this solution alot, I thought some thing like this was possible, but my Rust-foo is not strong enough and I have a hard time navigating the documentation to dig up the idiomatic Rust examples

.

Use integers as soon as possible -> no rounding needed... Cheers.
PS: note that more than 10000 guys passed the kata.