Clojure reference Project up to date?

Question

Starting with Clojure I discovered a talk by Rich Hickey where he demonstrates some of Clojure's strengths on a basic Ant-Simulator.

Can this code still be considered as a good reference for Clojure? Especially the parts when he recursively sends off functions to agents to simulate a game loop. Example:

(defn animation [x]
  (when b/running
    (send-off *agent* #'animation))
    (. panel (repaint))
  (. Thread (sleep defs/animation-sleep-ms))
  nil)

Edit:

I am not interested in the #' reader macro but more wether it is idiomatic/good Clojure to recursively call a function on a agent or not.

Answer 1

This snippet is current in Clojure 1.4. Is it idiomatic for a function to submit a task back to the agent that called it? Yes.

Here is an example that uses a similar approach to recursively calculate a factorial:

(defn fac [n limit total]
  (if (< n limit)
    (let [next-n (inc n)]
       (send-off *agent* fac limit (* total next-n)) 
       next-n)
     total))

 (def a (agent 1))

 (await (send-off a fac 5 1))
 ; => nil
 @a
 ;=> 120

Update

The above is a contrived example and actually not a good one, as there is a race condition between the various recursive send-off calls and the later await. There may be some send-off calls yet to be added to the agent's task queue.

I re-wrote the above as follows:

(defn factorial-using-agent-recursive [x]
  (let [a (agent 1)]
    (letfn [(calc  [n limit total]
               (if (< n limit)
                 (let [next-n (inc n)]
                   (send-off *agent* calc limit (* total next-n))
                   next-n)
                 total))]
      (await (send-off a calc x 1)))
    @a))

and observed the following behavior:

user=> (for [x (range 10)] (factorial-using-agent-recursive 5))
(2 4 3 120 2 120 120 120 120 2)
user=> (for [x (range 10)] (factorial-using-agent-recursive 5))
(2 2 2 3 2 2 3 2 120 2)
user=> (for [x (range 10)] (factorial-using-agent-recursive 5))
(120 120 120 120 120 120 120 120 120 120)

Moral of the story is: don't use agents for synchronous calculations. Use them for asynchronous independent tasks - like updating animations displayed to a user :)