I've found in that case it's much easier to just operate on raw bytes, then transform those into UTF characters. It works trivially for UTF8 and needs some mass

I've found in that case it's much easier to just operate on raw bytes, then transform those into UTF characters. It works trivially for UTF8 and needs some massaging for UTF16 and UTF32 because BE/LE.