The [0..n] slicing is what I was surprised to not see being done. As for what you're suggesting, using from_utf8 adds a overhead to check that you're giving it

The [0..n] slicing is what I was surprised to not see being done. As for what you're suggesting, using from_utf8 adds a overhead to check that you're giving it valid UTF-8 (this does not get optimized away). Dipping into unsafe can help here: const INDENTS: [u8; 256] = [b' '; 256]; // ... IndentStyle::Spaces(n) => unsafe { std::str::from_utf8_unchecked(&INDENTS[0..n as usize]) }, But I would consider having INDENTS be a &str to get the cheap slicing without needing unsafe. Since putting in a string literal of 256 spaces is nasty I would use the const_format crate to generate the constant: const INDENTS: &str = str_repeat!(" ", 256);