Clojure isn't doing the caching. The JVM is doing the caching. In this case, Clojure just has mechanical sympathy for how the JVM operates.
On Thursday, January 22, 2015 at 11:10:56 PM UTC-5, Michael Blume wrote: > > It sounds like basically dispatch is fast because we bothered to make it > fast (by caching) and satisfies? is slow because we didn't. Is it worth > throwing caching at satisfies? to make it fast or should satisfies? just > not be on the critical path for Clojure code? > > (To give a motivating example, satisfies? is on the critical path for > honeysql and will be until this pull is merged: > https://github.com/jkk/honeysql/pull/38) > > > On Thu Jan 22 2015 at 5:52:13 PM Ghadi Shayban <gsha...@gmail.com > <javascript:>> wrote: > >> Protocol call sites build an inline cache to optimize dispatch. The >> benchmark is running many times and reaping benefit from the cache. >> satisfies? looks up the object's class in the protocol's implementation >> map [1], and the benchmark is stressing this. You'll see that code checks >> if the protocol has the backing interface first, then checks for the >> object's class, then if necessary walks up the superclass chain. >> >> [1] >> https://github.com/clojure/clojure/blob/master/src/clj/clojure/core_deftype.clj#L507-L516 >> >> >> >> On Thursday, January 22, 2015 at 8:36:23 PM UTC-5, Michael Blume wrote: >> >>> Extends seems to be defeated by superclassing. ie: >>> >>> (extends? my-protocol (class {})) => false >>> >>> because my-protocol is extended to IPersistentMap and (class {}) isn't >>> IPersistentMap it's PersistentArrayMap (which implements IPersistentMap but >>> extends? doesn't care) >>> >>> On Thu Jan 22 2015 at 5:28:30 PM Timothy Baldridge <tbald...@gmail.com> >>> wrote: >>> >> The logic of extends? is much simpler, so try that. IIRC it's something >>>> like "extends? returns true if any method on the protocol is implemented >>>> by >>>> x", "satisfies? returns true if all methods of a protocol are implemented >>>> by x". >>>> >>>> The docs don't seem to give much help here, so play with it in the repl >>>> a bit. >>>> >>>> Timothy >>>> >>>> On Thu, Jan 22, 2015 at 6:14 PM, Michael Blume <blume...@gmail.com> >>>> wrote: >>>> >>>>> (defprotocol my-protocol >>>>> (foo [this])) >>>>> >>>>> (extend-protocol my-protocol >>>>> clojure.lang.IPersistentMap >>>>> (foo [this] "hello from map")) >>>>> >>>>> (criterium.core/quick-bench >>>>> (satisfies? my-protocol {})) >>>>> >>>>> (criterium.core/quick-bench >>>>> (foo {})) >>>>> >>>>> Simply calling foo on an empty map takes 7 ns, >>>>> but checking whether the map satisfies my-protocol takes 22 µs, 3000 >>>>> times longer. >>>>> >>>>> It seems like to call foo, some mechanism has to look up an >>>>> implementation of my-protocol for maps -- how is it we can do that so >>>>> quickly for a call and so slowly for satisfies? >>>>> >>>>> -- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "Clojure" group. >>>>> >>>> To post to this group, send email to clo...@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+u...@googlegroups.com >>>> >>>> >>>>> For more options, visit this group at >>>>> http://groups.google.com/group/clojure?hl=en >>>>> --- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "Clojure" group. >>>>> >>>> To unsubscribe from this group and stop receiving emails from it, send >>>>> an email to clojure+u...@googlegroups.com. >>>> >>>> >>>>> For more options, visit https://groups.google.com/d/optout. >>>>> >>>> >>>> >>>> >>>> -- >>>> “One of the main causes of the fall of the Roman Empire was >>>> that–lacking zero–they had no way to indicate successful termination of >>>> their C programs.” >>>> (Robert Firth) >>>> >>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "Clojure" group. >>>> >>> To post to this group, send email to clo...@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+u...@googlegroups.com >>> >>> >>>> For more options, visit this group at >>>> http://groups.google.com/group/clojure?hl=en >>>> --- >>>> You received this message because you are subscribed to the Google >>>> Groups "Clojure" group. >>>> >>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to clojure+u...@googlegroups.com. >>> >>> >>>> For more options, visit https://groups.google.com/d/optout. >>>> >>> -- >> You received this message because you are subscribed to the Google >> Groups "Clojure" group. >> To post to this group, send email to clo...@googlegroups.com >> <javascript:> >> Note that posts from new members are moderated - please be patient with >> your first post. >> To unsubscribe from this group, send email to >> clojure+u...@googlegroups.com <javascript:> >> For more options, visit this group at >> http://groups.google.com/group/clojure?hl=en >> --- >> You received this message because you are subscribed to the Google Groups >> "Clojure" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to clojure+u...@googlegroups.com <javascript:>. >> For more options, visit https://groups.google.com/d/optout. >> > -- 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 --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
