Collecting unique values from multiple columns in Cascalog. https://groups.google.com/forum/#!topic/cascalog-user/CjpApzUiwHw

gistfile1.clj

(defparallelagg collect-set
                :init-var (mapfn [s] #{s})
                :combine-var into
                :present-var identity)

 (let [set->string (mapfn [s1 s2] (clojure.string/join "," (clojure.set/union s1 s2)))]
  (??<- [?id ?fruits]
   ([["1" "banana" "grape"]
     ["1" "apple" "apple"]
     ["1" "apple" "lemon"]
     ["2" "orange" "kiwi"]
     ["2" "kiwi" "apple"]] ?id ?fruit1 ?fruit2)
   (collect-set ?fruit1 :> ?fruits1)
   (collect-set ?fruit2 :> ?fruits2)
   (set->string ?fruits1 ?fruits2 :> ?fruits)))

;;=> (["1" "apple,banana,grape,lemon"] ["2" "apple,orange,kiwi"])