Very well, something along these lines would be my guess too. But that would mean that in case 2 protocols are no faster than multimethods. And I've got an impression that protocols are described to be as fast as interface dispatch (callvirt). So either my impression is wrong (which is totally possible) or there is some clever trick available on JVM to do case 2 below faster. That's why I'm asking.
Also for case 1 your description implies that even for your own types there is a performance penalty for extending interfaces *after* type definition. I'm curious if that's the case as well. - Dmitry On Saturday, May 7, 2011 11:29:16 PM UTC-7, Krukow wrote: > When you extend the protocol to reach a type there are two cases > > 1) if you are simultaneously defining a new type and extending the > protocol, the underlying new class of the type can directly implement > the interface. The dispatch function simply calls the appropriate > interface method directly on the object (first argument of the > protocol function). > > 2) (your case) you are extending an existing type. This dispatch > function is changed to take account if this new case (I believe the > dispatch is a switch on the class of the first argument to the > protocol function). > > -- 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 Note that posts from new members are moderated - please be patient with your first post. 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
