Thank you both Matthias and Roman for explaining those things so thoroughly :)
Matthias Felleisen <matthias@...> writes: > Second, here is why your code is potentially buggy. Suppose you used the type of your mystringlistlist as an > argument type to f. You could then apply f to '( () ), which is a list of list of strings. If this argument flows > to assoc, you violate the primitives basic constraint, namely, that each item on the list consists of at > least one cons cell. By saying (cons String (Listof String)) you are saying list of non-empty list of > strings, which fits the assoc type signature fine. > > Third, the reason your code doesn't have a run-time error -- i.e. the type error is conservative and the > 'potential' is needed -- is because your specific list consists of non-empty lists of strings. Even > though TR can infer this type, your type assignment to this identifier overrides TR's capabilities. Oh, I really didn't think that far here... I guess TR is a bit more strict than I imagined. Which is a good thing, of course. I'd rather have some headaches declaring my functions and types correctly than having to debug through layers of old code just because some symbol went into some list where it shouldn't have been. But I see that I have to be more careful to prevent runtime errors in the future. Best Regards, Tim ____________________ Racket Users list: http://lists.racket-lang.org/users
