The tokenizer gives us whole float literal tokens, we have to split them up in order to be able to create field access from them.