Oh, I fully agree. I was more interested in the benchmark result. I really expected map to be much more expensive than slicing so I was rather surprised by the result. The result holds good for arrays of at least 100_000 elements.