This is interesting. The `Any` type in Typed Racket includes values that may have higher-order "original" types. For example, it may have originally been a typed function (-> Fixnum Fixnum) or (-> String Boolean), such that if you call it without a contract guarding it, it should be an error. This is because `Any` is a super type of those types.
However, the value returned by `dynamic-require` is different from a normal `Any`. It will never be an unguarded "originally" typed value, so it should be safe to pass it to `cast` without adding a new guard. It might make sense to define a new type `UntypedAny` or maybe `GuardedAny`. There are two differences: (1) this type is *not* a super type of any other type that might involve higher-order values (2) it is safe to pass a value of this type into untyped code without an additional guard Point (2) is what would allow `cast` to work, and point (1) is what would make that safe to do. With this new `UntypedAny` type, dynamic-require could return the type `UntypedAny` instead of `Any`. -------------------------------------------------- Why this would allow `cast` to work (and why `require/typed` is a good workaround for now) The difference between `cast` and `require/typed` is this: - `require/typed` guards value going from Untyped -> Typed - `cast` guards values going from Typed -> Untyped -> Typed The `cast` needs the extra Typed -> Untyped boundary to guard "originally" typed higher-order values, such as a (-> Fixnum Fixnum) function annotated as `Any`. However, `require/typed` takes its values from an untyped context to begin with, so it doesn't need to deal with that extra boundary. The value returned by `dynamic-require` is never one of those typed higher-order values, and if the type system knows that through the `UntypedAny` type, the type system will know it is safe to make `cast` behave like `require/typed`. It will know it is safe to pass it only through the Untyped -> Typed boundary. Alex Knauth > On Apr 16, 2018, at 10:01 AM, Matthias Felleisen <matth...@felleisen.org> > wrote: > > > Have you considered organizing your program as follows: > > #lang racket > > (module a racket > (provide f) > (define (f x) 10)) > > (module b racket > (provide g) > (define g (dynamic-require '(submod "foo.rkt" a) 'f))) > > (module c typed/racket > (require/typed (submod ".." b) [g (-> Integer Integer)]) > (g 20)) > > (require 'c) > > That is, run the dynamic-require in an untyped module and then import it at > the desired type? > > > >> On Apr 16, 2018, at 5:22 AM, mailoo <dmichi...@mailoo.org >> <mailto:dmichi...@mailoo.org>> wrote: >> >> >> Hello, >> >> I'm new to racket, and even more with typed/racket. >> >> I play a little with the "Any" type (due to 'dynamic-require' which return >> Any), and I'm not able to cast them back in a function. >> >> I (over) simplify my question with this little program : >> >> ``` >> (: p Any) >> (define (p i) (displayln i)) >> >> ; Here I want to get back my function >> (define proc (cast p (-> Integer Void))) >> (proc 2) >> >> ``` >> >> but I get this error when I try to execute the function : >> >> ``` >> ; contract violation >> ; Attempted to use a higher-order value passed as `Any` in untyped code: >> #<procedure:p> >> ; in: Any >> ; contract from: typed-world >> ; blaming: cast >> ; (assuming the contract is correct) >> ; at: <pkgs>/racketFBP/test.rkt:13.13 >> ; Context: >> ; /home/denis/dev/racket/racketFBP/test.rkt:1:1 [running body] >> ; [Due to errors, REPL is just module language, requires, and stub >> definitions] >> ``` >> >> In reality, I get my function by `(dynamic-require path 'p)`, and I didn't >> find a typed version of this function... >> >> Is there a way to go from Any to a function, or to replace my use of >> `dynamic-require`? >> >> Thank you in advance, >> Denis Michiels >> >> -- >> 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 racket-users+unsubscr...@googlegroups.com >> <mailto:racket-users+unsubscr...@googlegroups.com>. >> For more options, visit https://groups.google.com/d/optout >> <https://groups.google.com/d/optout>. > > > -- > 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 racket-users+unsubscr...@googlegroups.com > <mailto:racket-users+unsubscr...@googlegroups.com>. > For more options, visit https://groups.google.com/d/optout > <https://groups.google.com/d/optout>. -- 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 racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
