Bank ocr

gistfile1.clj

(require '[clojure.string :as str])

(def characters (apply str
  " _     _  _     _  _  _  _  _ \n"
  "| |  | _| _||_||_ |_   ||_||_|\n"
  "|_|  ||_  _|  | _||_|  ||_| _|\n"
  "                              \n"))

(defn chunk3 [line]
  (->> (partition 3 line)
       (map #(apply str %1))))

(defn transpose [lst]
  (apply mapv vector lst))

(def all-numbers
   (->> characters
      (str/split-lines)
      (map chunk3)
      (transpose)
      (map-indexed vector)
      (map (fn[[a b]] [b a]))
      (into {})))

(defn ocrToNumber [ocr]
  (->> ocr
       (str/split-lines)
       (map chunk3)
       (transpose)        
       (map #(get all-numbers %1))))

(def toOcr (apply str 
  "    _     _ \n"
  "  ||_||_||_ \n"
  "  ||_|  ||_|\n"
  "            \n"))

(println toOcr)
(println (ocrToNumber toOcr))

No, because (partition 3 line) returns a seq of lazyseq (of chars):

(map str (partition 3 "abcxyz"))
("clojure.lang.LazySeq@1ecc1" "clojure.lang.LazySeq@245f8")

Partition is a general purpose function for seqs, and strings are just char seqs. I suppose there might exist a better partition function.

Ah, ok. What's kind of incredible is that if I apply the change I mentioned and run the program, I still find that it actually still works as expected. Maybe the stringified versions of the lazy sequences are stable due to caching/memoization. If so, that's pretty impressive!

What version of clojure are you using? I am using 1.5.1 and if I apply your change all I get is (nil nil nil nil).

It's 1.5.1 on OS X.