Make random-fodder-seq thread safe

- The details of the bug and fix are in the docstring of
`hyperion.key/random-fodder-seq`

api/spec/hyperion/key_spec.clj

@@ -47,6 +47,15 @@
     (should= true (seq? (random-fodder-seq)))
     (should= 10 (count (set (map (fn [_] (take 5 (random-fodder-seq))) (range 10))))))
 
+  (it "random-fodder-seq is threadsafe"
+     (should= 250
+        (reduce +
+                (map
+                  (fn[_] (count (set (pmap
+                                       (fn [_] (take 10 (random-fodder-seq)))
+                                       (range 50)))))
+                  (range 5)))))
+
 ;  (it "generate key times"
 ;    (prn (take 100 (iterate (fn [_] (generate-id)) nil)))
 ;    (prn (take 100 (iterate (fn [_] (generate-id2)) nil)))

api/src/hyperion/key.clj

@@ -23,7 +23,13 @@
 (def fodder-count (count key-fodder))
 
 (defn random-fodder-seq
-  ([] (random-fodder-seq (java.util.Random. (System/nanoTime))))
+  "The default generator for `random-fodder-seq` is threadsafe for JDK 1.6b73
+  or greater (http://bugs.java.com/view_bug.do?bug_id=6379897). Calls to this on
+  multiple threads (i.e. `(pmap (fn [_] (take 10 random-fodder-seq)) (range 10))`)
+  will have a different seed for each instance  of `java.util.Random`. If on a JDK
+  earlier than JDK 1.6b73 there is a chance that there will be multiple instances
+  of `java.util.Random` with the same seed causing collisions"
+  ([] (random-fodder-seq (java.util.Random.)))
   ([generator]
     (cons
       (aget key-fodder (.nextInt generator fodder-count))