I think I may have figured it out. New patch attached to ticket CLJ-1065 that should eliminate run-time checks for duplicate map keys, for those maps whose keys are all compile-time constants.
Andy On Sep 8, 2012, at 4:38 PM, Andy Fingerhut wrote: > Rich: > > I'm not sure what you mean by the not-fastest-path possible that exists in > today's Clojure code, so if you get a chance, see if the below is what you > mean. > > As far as I can tell (i.e. putting debug println's in the Java code of > RT.map), when someone enters a map literal in, say, a function definition, > and all keys *and* values are compile time constants, it calls RT.map() while > the function is being compiled, but never again when the function is called. > > If I make a similar function with run-time variable keys or values, RT.map() > is called every time the function is invoked. Each of these calls repeats > the check that the keys are unique. > > Do you mean that you want a new code path where if the keys are compile time > constants, but the values are variables at compile-time, then at run time > this map should be created with a method that avoids the unnecessary check > for unique keys? > > And by the word "restore" do you mean to imply that it was this way at one > time before? > > Thanks, > Andy > > > On Sep 8, 2012, at 5:29 AM, Rich Hickey wrote: > >> Thanks! >> >> I'm still interested in patch for recommendation #3: >> >> Restore the fastest path possible for those cases where the keys are >> compile-time detectable unique constants >> >> I'd like to see all three recommendations go into a release as a set. >> >> >> On Sep 8, 2012, at 2:22 AM, Andy Fingerhut wrote: >> >>> The new ticket CLJ-1065 has a patch that I think implements the desired >>> behavior on the dev wiki page. >>> >>> i.e. set/map literals with duplicates are invalid (status quo) >>> >>> All constructor functions for sets and maps allow duplicates, and for maps, >>> always take the value associated with the last occurrence of the same key. >>> All constructor functions explicitly say this in their doc strings. >>> >>> Andy >>> >>> On Sep 7, 2012, at 2:06 PM, Rich Hickey wrote: >>> >>>> >>>> On Sep 7, 2012, at 3:35 PM, Sean Corfield wrote: >>>> >>>>> On Fri, Sep 7, 2012 at 10:49 AM, Rich Hickey <richhic...@gmail.com> wrote: >>>>>> I've added my feedback there >>>>>> (http://dev.clojure.org/display/design/Allow+duplicate+map+keys+and+set+elements) >>>>> >>>>> Thanx Rich! So the recommendation is: >>>>> >>>>> * set/map literals with duplicates are invalid (status quo) >>>>> >>>>> * hash-set/hash-map should change (to last key wins, as if conj'd/assoc'd) >>>>> >>>>> * sorted-set/sorted-map should not change (last key wins, as if >>>>> conj'd/assoc'd) >>>>> >>>>> * array-map should not change (throws on dupes)? >>>>> >>>>> Highlighting that last one since it's not mentioned on the wiki and >>>>> would then be the "odd one out" but perhaps there's a good reason? >>>> >>>> No, array-map should be the same too. >>> >>> -- >>> 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 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 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
