Could you try the numeric keys against the master branch? The num/prim/equiv changes should help with this.
Stu > Interesting possibility, hard for me to test for directly. But > following that idea, I changed the keys to strings and the problem > went away, so I'm happy :) Thanks! > > > > On Tue, Sep 14, 2010 at 4:45 AM, Stuart Halloway > <stuart.hallo...@gmail.com> wrote: >> Can you look at the class of these keys and see if some are Long while >> others are Integer? >> >> Stu >> >>> I'm running a large hadoop job in which merge-with is called millions >>> of times to aggregate values among about 1000 keys. Basically we are >>> counting the number of times the keys occur among all entries and >>> using merge-with as the reduce function. >>> >>> In the output, the keys are often duplicated (I've attached a sample >>> of the result map below so you can see what I mean, the keys 14336, >>> 14464, 14528 have dups), with the repeat having a value of 1 or >>> sometimes 2. >>> >>> This is coming right out of merge-with, so its baffling why there >>> would be duplicate keys, unless there is some kind of bug in that >>> function when called in a strenuous setting. Another theory is >>> something going wrong during serialization in intermediate steps. I am >>> using cascading-clojure. Thanks for your input. >>> >>> {14336 18176, 14336 2, 14368 19111, 14400 17161, 14432 17015, 14464 >>> 14604, 14464 1, 14496 20810, 14528 18759, 14528 1, 14560 28086, 13568 >>> 162, 14592 31956, 13600 437, 14624 38402, 13632 281, 14656 38429, >>> 13664 1137, 14688 29531} >>> >>> -- >>> 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 -- 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