> On May 27, 2016, at 4:24 PM, 'John Clements' via Racket Users > <[email protected]> wrote: > > I have a funny case in TR where occurrence typing loses information about the > type of a term: > > #lang typed/racket > > (: expand (Sexp -> Sexp)) > (define (expand s) > (match s > [(list (? symbol? id) a ...) > s])) > > This fails to type check, because the occurrence of ’s’ in the rhs of the > match is no longer known to be an Sexp.
In the latest snapshot versions, this type-checks. It might be because of the addition of Andrew Kent's intersection types, because according to DrRacket the type of the `s` in the body is: (Pairof Symbol (U Null (Pairof Sexp (∩ Sexp (Listof Any))))) which is a subtype of Sexp. For previous versions, a workaround I've used is to add an extra `my-identity` function to fool it it so that it doesn't realize that I'm applying a predicate to `s` or matching on it. > Specifically, I get the message: > > Type Checker: type mismatch > expected: Sexp > given: (Pairof > Symbol > (U Null (Pairof Sexp (U Null (Pairof Sexp (Listof Any)))))) in: s > > I understand that the (Pairof …) type arises from the expansion of match, but > it seems unfortunate that TR can’t compute the intersection of the two types > known to hold for ’s’ to arrive at > > (Pairof Symbol (U Null (Pairof Sexp (U Null (Pairof Sexp (Listof Sexp)))))) > > … which is a subtype of Sexp, and therefore legal to return as an Sexp. > > Is there perhaps a hinting mechanism by which I could suggest to TR that it > look at the original type known to be associated with ’s’? Or perhaps there’s > a depth limit on the intersection computation that I’m running into? Or > perhaps I’m misunderstanding the situation entirely? > > John > > p.s.: I can obviously work around this problem with a ‘cast’. > > [*] “weakens” may not be an appropriate term here, given that neither of the > two types discussed are subtypes of the other. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
