When to use loop/recur in Clojure?
When you first start learning Clojure, you’ll eventually come across the loop/recur construct that enables you to essentially do tail recursion.
Why does Clojure need to have loop/recur when you can simply use a function and call it again?
Let’s say you’re writing a tail-recursive (iterative) function that calculates the nth fibonacci number.
Since we’re recursing from the tail position, what’s the difference between calling the function by its name:
(defn fib [prev cur n]
(if (zero? n)
prev
(fib cur (+ prev cur) (dec n))))
user> (fib 0 1 6)
8
and using recur?
(defn fib-recur [prev cur n]
(if (zero? n)
prev
(recur cur (+ prev cur) (dec n))))
user> (fib-recur 0 1 6)
8
Note that I’m using recur without using loop. If you don’t use loop, recur will use the function in the lexical context as the recur target instead. loop is just a macro that helps you make a recur target. It’s like making another function but with some sugar to make it a bit nicer. We could have written the fibonacci function with loop instead of the function trivially
(loop [prev 0
cur 1
n 6]
(if (zero? n)
prev
(recur cur (+ prev cur) (dec n))))
8
The biggest difference is that using recur you won’t blow the stack for large numbers of recurisve calls:
;; The M that follows the number indicates a bigdecimal
user> (fib 0M 1M 10000M)
StackOverflowError clojure.lang.PersistentHashMap$BitmapIndexedNode.index (PersistentHashMap.java:677)
Versus the exact same algorithm using the recur:
user> (fib-recur 0M 1M 10000M)
33644764876431783266621612005107543310302148460680063906564769974680081442166662368155595513633734025582065332680836159373734790483865268263040892463056431887354544369559827491606602099884183933864652731300088830269235673613135117579297437854413752130520504347701602264758318906527890855154366159582987279682987510631200575428783453215515103870818298969791613127856265033195487140214287532698187962046936097879900350962302291026368131493195275630227837628441540360584402572114334961180023091208287046088923962328835461505776583271252546093591128203925285393434620904245248929403901706233888991085841065183173360437470737908552631764325733993712871937587746897479926305837065742830161637408969178426378624212835258112820516370298089332099905707920064367426202389783111470054074998459250360633560933883831923386783056136435351892133279732908133732642652633989763922723407882928177953580570993691049175470808931841056146322338217465637321248226383092103297701648054726243842374862411453093812206564914032751086643394517512161526545361333111314042436854805106765843493523836959653428071768775328348234345557366719731392746273629108210679280784718035329131176778924659089938635459327894523777674406192240337638674004021330343297496902028328145933418826817683893072003634795623117103101291953169794607632737589253530772552375943788434504067715555779056450443016640119462580972216729758615026968443146952034614932291105970676243268515992834709891284706740862008587135016260312071903172086094081298321581077282076353186624611278245537208532365305775956430072517744315051539600905168603220349163222640885248852433158051534849622434848299380905070483482449327453732624567755879089187190803662058009594743150052402532709746995318770724376825907419939632265984147498193609285223945039707165443156421328157688908058783183404917434556270520223564846495196112460268313970975069382648706613264507665074611512677522748621598642530711298441182622661057163515069260029861704945425047491378115154139941550671256271197133252763631939606902895650288268608362241082050562430701794976171121233066073310059947366875M
You pretty much almost always want to use loop/recur over tail recursively calling the function when you can for this reason.
Join the 80/20 DevOps Newsletter
If you're an engineering leader or developer, you should subscribe to my 80/20 DevOps Newsletter. Give me 1 minute of your day, and I'll teach you essential DevOps skills. I cover topics like Kubernetes, AWS, Infrastructure as Code, and more.
Not sure yet? Check out the archive.
Unsubscribe at any time.