Very close to some sort of "Type Classes" for structures if a) Ability to define and label a set of typed generic methods. b) Declare a structure as satisfying a given set of typed generic methods. c) TR verify said satisfaction.
I've successfully attached a TR typed writer method to a struct:'s #methods gen:custom-write for example. On Tue, Aug 14, 2012 at 10:17 AM, Asumu Takikawa <[email protected]> wrote: > On 2012-08-13 13:28:07 -0700, Jordan Johnson wrote: > > I'm curious: what use cases call for the new generics as opposed to > > the old class & interface system? I'm trying to get a clearer picture > > of the motivation for the new functionality, and the differences > > between old-style class-based OO code and this struct-based kind. > > This feature isn't intended to replace class-based OO code. Notably, > these generics don't have inheritance, overriding, mixins, and other > "heavyweight" features that you might use with the class system. > > Instead, racket/generic is intended to replace the use of struct type > properties as a mechanism to add generic behavior to datatypes. Before > this library, generic functions like `dict-ref` or `stream-first` relied > on ad-hoc protocols where a struct would provide a collection of > functions to implement dictionary operations. > > Now we provide a unified and consistent way to define such generic > functions and implement their behavior on specific instances. > > Cheers, > Asumu > ____________________ > Racket Users list: > http://lists.racket-lang.org/users >
____________________ Racket Users list: http://lists.racket-lang.org/users
