Why Clojure? Part 2: Effortless async by design
As a second part in what I’ve decided will be a series attempting to justify my choice of Clojure as a general-purpose web language, I’d like to talk a little bit about writing asynchronous code in Clojure, and just how easy the design decisions made by Mr. Hickey make that.
Clojure web handling code works in the pretty standard way; you write a function that accepts a request object and returns a response object, and whatever happens in between is up to you. In many cases, though, the in-between part is something that doesn’t have an impact on your response; a good example (which I will be referring to constantly) is sending an email in response to a user action.
The basics of async with Clojure are the same as Java; Clojure functions implement Java’s Runnable interface, and so can be spun off as regular Java threads with no extra effort.
(defn do-a-thing  ; A thing you want to do ) (defn do-something-else  ; Another thing you want to do ) (defn do-two-things  (-> (Thread. do-a-thing) .start) (-> (Thread. do-something-else) .start)
You can make your threads closures to pass values to them:
(defn do-something-with-a-request [req] (fn  ; Use the state somehow )) (defn handle-request [req] (-> (Thread. (do-something-with-a-request req)) .start) ; Carry on as normal )
Is your app sending email as part of a synchronous request? It shouldn’t be. Are you logging requests as part of a synchronous request? You shouldn’t be. Asyncing these activities in other popular languages could be considered premature optimization; in Clojure it’s just wrapping a function.
Just remember that your threads made this way should really be keeping to themselves. If you need to access shared state, you’d better make sure you have a thread-safe way of doing it.
If you will need the computed value later but not immediately, you can use
future. This will yield a future reference that, when dereferenced with
deref, will give you the value of the computation, blocking if necessary.
Here’s another pointless example:
(defn do-something-with-a-request [req] ; Implementation ) (defn do-something-with-result [result] ; Something else ) (defn handle-request [req] (let [result (future (fn  (do-something-with-a-request req)))] ; Do something else (do-something-with-result @result)))
Did someone say thread-safe?
I dare you to find a thread-safer language. Well, maybe Erlang.
I wasn’t able to produce a better, simpler explanation of the differences between them than this Stack Overflow response, so I’ll just quote it here.
- Refs are for coordinated, synchronous access to many identities".
- Atoms are for uncoordinated, synchronous access to a single identity.
- Agents are for uncoordinated, asynchronous access to a single identity.
- Vars are for thread local isolated identities with a shared default value.
Coordinated access is used when two identities need to be changes together, the classic example being moving money from one bank account to another; it needs to either move completely or not at all.
Uncoordinated access is used when only one identity needs to update, this is a very common case.
Synchronous access is where the call expects to wait until all the identities are settled before continuing.
Asynchronous access is “fire and forget” and lets the identity reach its new state in its own time.
Any of these should, of course, be used sparingly. If you can avoid state, do so.
Lamina: Advanced asynchronous processing
Lamina is a Clojure library that makes passing values via a queue-like interface super-easy. Like, crazy easy. Here’s an example that prints “Hello World” every second using an asynchronous queue.
(ns example.helloqueue (:require [lamina.core :as lamina])) (def ch (lamina/channel)) (defn loop-forever [f] ; Since repeatedly is lazy, we wrap it in doall (doall (repeatedly f))) (defn producer  (loop-forever (fn  (lamina/enqueue ch "Hello World!") (Thread/sleep 1000)))) (defn consumer  (loop-forever (fn  (println @(lamina/read-channel ch))))) (defn main  (-> (Thread. producer) .start) (-> (Thread. consumer) .start))
What happened there?
lamina/channel defines a channel.
sends a message to it, and
lamina/read-channel reads from it. When you
read-channel, it blocks until a message appears in the
Update: @ztellman, Lamina’s author, recommends
to streamline this example by dispensing with the
producer function entirely:
(ns example.helloperiodically (:require [lamina.core :as lamina])) (def ch (lamina/periodically 1000 (fn  "Hello World!"))) (def loop-forever (comp doall repeatedly)) (defn consumer  (loop-forever (fn  (println @(lamina/read-channel ch))))) (defn main  (-> (Thread. consumer) .start))
Lamina is a big library with tons of good stuff in it, so don’t take this as any sort of comprehensive example.
Hello World too trivial? How about sending notification emails from your app:
(ns example.mailqueue (:require [lamina.core :as lamina])) (defn send-mail [messagedata] ; Implement as required ) (def mail-queue (lamina/channel)) (defn enqueue-mail [messagedata] ; Called whenever you want to send an email (lamina/enqueue mail-queue messagedata)) (defn consume-mail  (let [messagedata @(lamina/read-channel mail-queue)] (send-mail messagedata))) (defn mail-consumer  (doall (repeatedly consume-mail)) ; Be sure to start the mail-consumer thread when you start your app (defn main  (-> (Thread. mail-consumer) .start))
If you can’t be bothered to remove the overhead of sending email from your main request processing for the cost of 20 lines of code, well, that’s just too bad.
And that’s just for starters
There are so many different ways to use the tools the Clojure provides, and that others have built on them, to improve your application. I haven’t been using the language for all that long, but I wanted to share a few of the ways it’s helped me.
Try out Clojure today! You’ll be amazed what you learn.