comparison.clj

;; benchmark 1
;;
user> (time (count in-seq))
"Elapsed time: 2225 msecs"
812990
;;
;; memoizing
;;
user> (time (count in-seq))
"Elapsed time: 208 msecs"
812990
user> (time (count in-seq))
"Elapsed time: 213 msecs"
812990
;;
user> (/ 2225 812990)
0.00273681103088599


;; benchmark 2, CLR version
;;
;; from Clojure Data Analysis Cookbook
;; the payload function - here freq. - could be trivially factored out,
;; a let binding in lazy-read-ok could keep a global state for it.
;;
(defn lazy-read-ok
  "This one's OK. It handles everything lazy, but the lazy read is emeshed with
  the logic.

  This contrived example pulls the age out of each file.

  Note that you still need to force everything to be read. Here I'm using
  frequencies to do this."
  ([csv-file]
   (with-open [in-file (io/text-reader csv-file)]
     (frequencies (map #(nth % 2) (csv/read-csv in-file :separator \tab))))))

user=> (time (lazy-read-ok in-file))
"Elapsed time: 99062 msecs"
{"2001-02-12" 1640, "2001-03-02" 1646, "2001-01-22" 1639, "2000-03-03" 1558, "2001-12-13" 1731, "2001-04-03" 1653, "2000-06-22" 1568, "2001-08-29" 1704, "2001-11-23" 1720, "2000-12-12" 1637, ...}
user=> (/ 99062.0 812990)
0.12184897723219228