Even YARD has some docs that use literal ruby in the type annotation like so
where we should parse :root as "literally the symbol :root".
I think if parsing a type fails then we should try using Parser on the documented type. I think we should still reraise though if it isn't one of the very simple literals like sym, str, int, float, array, etc. Obviously I am not going to parse and eval arbitrarily complex ruby