Here is a simple example:

pub fn flatten_array(ys: Vec<[u32; 288]>) -> Vec<u32> { ys .into_iter() .flatten() .collect() }

Godbolt

It is obvious that the size and alignment will be exactly the same after conversion, yet right now Rust generates a bunch of code instead of efficiently reusing allocation like it does with similar vec.into_iter().map().collect() cases.

I'm sure there is a reason for this that might be non-trivial, but I don't think this is completely unreasonable either.

Similar issue happens with .flat_map(), which I tried to use originally.

UPD: Just discovered Vec::into_flattened(), which does exactly what I need with a simple .flatten(), but still doesn't cover .flat_map() and thus not as generic.