clojure.core.async/timeout

View this page on ClojureDocs

Type: function Added: Clojure unknown Examples: 3 Runnable: 0

([msecs])

Returns a channel that will close after msecs

Examples

by runexec on 2014-10-01 . May have evaluation errors. ▶

(go-loop [seconds (atom 0) add-seconds! #(swap! seconds + %)] (println "Waiting 1 second") (<! (timeout 1000)) (add-seconds! 1) (println "Waiting 2 seconds") (<! (timeout 2000)) (add-seconds! 2) (println (format "Waited %s seconds" @seconds))) #<ManyToManyChannel clojure.core.async.impl.channels.ManyToManyChannel@67ddcb0e> ;; Waiting 1 second ;; Waiting 2 seconds ;; Waited 3 seconds

by runexec on 2014-10-01 . May have evaluation errors. ▶

(doseq [n (range 10) :let [i (-> n inc range rand-nth)]] (go (<! (timeout (* i 1000))) (println n))) nil 5 3 0 6 4 7 1 2 9 8

by loeschzwerg on 2021-11-22 . May have evaluation errors. ▶

;; A timeout channel is just a channel that will be autoclosed (def t (timeout 1000)) ;;=> #'user/t (poll! t) ;; before AND after timeout ;;=> nil (a/put! t 'foo) ;; before timeout => true (a/poll! t) => foo (a/poll! t) => nil (a/put! t 'foo) ;; after timeout => false ;; You COULD use a timeout-channel as a normal channel. ;; The preferred option however would be using alts!! (let [timout-chan (timeout 1000) comms-chan (chan 10 (map inc) #(log/error % "whoops")) ;; genuine channel [val port] (alts!! [timeout-chan some-other-chan])] (cond (= port timeout-chan) (throw (TimeoutException. "Tooooo late.")) (and (= port comms-chan) (int? val)) (log/infof "SMOOOoooth %d" val) :else (throw (IllegalStateException. "Well.. shit.")))))

Note by hlship

Never close! the channel returned by timeout -- this is an unwritten rule.

The implementation sometimes will use an existing timeout channel, if it is within a few ms of the desired timeout. Closing the channel will affect multiple otherwise unrelated CSPs that may be blocked/parked waiting for the channel to close.

In our application, we saw this as spurious "timed out after waiting X ms" messages when the logged timestamps didn't support this possibility.

Thanks to Ryan Neufeld for identifying this!

Note by raymcdermott

Might be obvious but this construct is preferred in go blocks over Thread/sleep as it does not block

---

Content from the matching ClojureDocs page, with authors credited on each contribution.