Can't some elements of the problem be solved with some form of predicate dispatching as proposed by Meikel? Predicate dispatching would allows us to use _anything_ as a type (i.e. structs themselves), as well as allowing user defined functions to do the matching instead of being limited to isa? and the global hierarchy. This of course has some performance implications I imagine... http://groups.google.com/group/clojure/browse_thread/thread/f8b1be403c927b03/438c21a80072a105?lnk=gst&q=hierarchy#438c21a80072a105
I have to say that I find the lack of a type system around structs to be fairly liberating. You can invent one if you need one, but you are not constrained by it. However, if somebody comes up with something like CLJOS as a set of handy macros, or extracts whatever parts they deem relevant, why not put that in clojure-contrib so that people can at least have something to use for the simple cases as pointed out by Konrad? On Thu, Jan 22, 2009 at 10:25 AM, Konrad Hinsen <konrad.hin...@laposte.net>wrote: > > On 22.01.2009, at 15:26, Rich Hickey wrote: > > > It's pretty easy to write a trivial struct system, much harder to > > address performance, interop, compilability, dynamicity etc > > constraints. > > Indeed. > > > As a simple case, if a defstruct is re-evaluated, will objects created > > after that be of the same 'type' as objects created before? > > I'd say no, which resolves the next question: > > > What if fields have been added/removed? > > Anyway, I don't think that re-evaluation is a serious issue, except > in interactive development sessions. The typical use would be > defining a struct once and then not touch it any more. > > > I'd prefer people experiment with libraries built on the existing > > facilities, with an open mind as to the possibilities of categorizing > > things other than by their structure. > > The added flexibility of multimethods is quite appreciable in non- > trivial situations. But at the moment there is no simple way to > handle simple situations. Using structs is straightforward and > familiar from other languages. It is certainly not *the* solution to > categorizing, but it's a simple one that is good enough for many > applications. > > Konrad. > > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
