| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
assoc-ing an update with the same key is hard to read. update is known and
idiomatic.
; avoid
(assoc coll :a (+ (:a coll) 5))
(assoc coll :a (+ (coll :a) 5))
(assoc coll :a (+ (get coll :a) 5))
; prefer
(update coll :a + 5)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 1.0 | 1.0 | 
A macro that nests an unquote-splicing in a macro with a & body can lead
to subtle hard to debug errors. Better to wrap the unquote-splicing in
a do to force it into 'expression position'.
; avoid
`(binding [max mymax] ~@body)
; prefer
`(binding [max mymax] (let [res# (do ~@body)] res#))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| false | true | false | 1.18.0 | 1.18.0 | 
When defining methods for a multimethod, everything after the dispatch-val is given directly to fn. This allows for providing a name to the defmethod function, which will make stack traces easier to read.
Examples:
; avoid (defmethod some-multi :foo [arg1 arg2] (+ arg1 arg2))
; prefer (defmethod some-multi :foo some-multi--foo [arg1 arg2] (+ arg1 arg2))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Checks for (/ x 1).
; avoid
(/ x 1)
; prefer
x
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 1.2.1 | 
map is lazy, which carries a performance and memory cost. dorun uses seq iteration to realize the entire sequence, returning nil. This style of iteration also carries a performance and memory cost. dorun is intended for more complex sequences, whereas a simple map can be accomplished with reduce + conj.
run! uses reduce which non-lazy and has no seq overhead.
; avoid
(dorun (map println (range 10)))
; prefer
(run! println (range 10))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1 | 0.1 | 
Using the Obj/staticMethod form maps the method call to Clojure's natural function position.
NOTE: This rule is disabled if lint/prefer-method-values is enabled to prevent conflicting diagnostics.
This rule is unsafe, as it can misunderstand when a symbol is or is not a class.
; avoid
(. Obj staticMethod args)
(. Obj (staticMethod) args)
; prefer
(Obj/staticMethod args)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1 | 1.15.2 | 
Using the .method form maps the method symbol to Clojure's natural function position.
NOTE: This rule is disabled if lint/prefer-method-values is enabled to prevent conflicting diagnostics.
This rule is unsafe, as it can misunderstand when a symbol is or is not a class.
; avoid
(. obj method args)
; prefer
(.method obj args)
| Name | Default | Options | 
|---|---|---|
| :chosen-style | :dot | :dot,:method-value | 
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 1.16.0 | 1.16.0 | 
It's an error to have duplicate case test constants.
; avoid
(case x :foo :bar :foo :baz)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1.119 | 0.1.119 | 
deftype and defrecord will throw errors if you define multiple fields
with the same name, but it's good to catch these things early too.
; avoid
(defrecord Foo [a b a])
; prefer
(defrecord Foo [a b c])
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1 | 0.1 | 
Avoid wrapping functions in pass-through anonymous function defitions.
This rule is unsafe, as it can misunderstand when a function is or is not a method.
; avoid
(fn [num] (even? num))
; prefer
even?
; avoid
(let [f (fn [num] (even? num))] ...)
; prefer
(let [f even?] ...)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Idiomatic if defines both branches. when returns nil in the else branch.
; avoid
(if (some-func) :a nil)
; prefer
(when (some-func) :a)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Idiomatic if-let defines both branches. when-let returns nil in the else branch.
; avoid
(if-let [a 1] a nil)
; prefer
(when-let [a 1] a)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Idiomatic if defines both branches. when-not returns nil in the truthy branch.
; avoid
(if (some-func) nil :a)
; prefer
(when-not (some-func) :a)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
if-not exists, so use it.
; avoid
(if (not x) y z)
; prefer
(if-not x y z)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
when-not already defines an implicit do. Rely on it.
; avoid
(if-not x (do (println :a) (println :b) :c))
; prefer
(when-not x (println :a) (println :b) :c)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Two nots cancel each other out.
; avoid
(if-not (not x) y z)
; prefer
(if x y z)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
or exists so use it lol.
; avoid
(if x x y)
; prefer
(or x y)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 1.19.0 | 
vec and set are succinct and meaningful.
; avoid
(into [] coll)
; prefer
(vec coll)
; avoid
(into #{} coll)
; prefer
(set coll)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1.69 | 0.1.69 | 
if-let exists so use it.
Suggestions can be wrong as there's no code-walking to determine if result binding is used in falsy branch.
; avoid
(let [result (some-func)] (if result (do-stuff result) (other-stuff)))
; prefer
(if-let [result (some-func)] (do-stuff result) (other-stuff))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1.69 | 0.1.69 | 
when-let exists so use it.
Suggestions can be wrong as there's no code-walking to determine if result binding is used in falsy branch.
; avoid
(let [result (some-func)] (when result (do-stuff result)))
; prefer
(when-let [result (some-func)] (do-stuff result))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | false | 1.16.0 | 1.16.0 | 
Synchronizing on interned objects is really bad. If multiple places lock on the same type of interned objects, those places are competing for locks.
; avoid
(locking :hello (+ 1 1))
; prefer
(def hello (Object.))
(locking hello (+ 1 1))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
loop has an implicit do. Use it.
; avoid
(loop [] (do (println 1) (println 2)))
; prefer
(loop [] (println 1) (println 2))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
Empty loops with nested when can be while. Doesn't apply if the final expr of the when isn't (recur), which includes any nested cases (let, etc).
; avoid
(loop [] (when (some-func) (println 1) (println 2) (recur)))
; prefer
(while (some-func) (println 1) (println 2))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | false | 1.20.0 | 1.20.0 | 
In interop scenarios, it can be necessary to add a type hint to mark a function's return type. This can be done by adding metadata to the function's name symbol or to the function's param vector. The former works but is prone to errors and is not recommended by the core team, whereas the latter is the official method. (See links below for further discussion.)
NOTE: Only checks defn forms. (Compare with eastwood's :wrong-tag linter.)
; avoid
(defn ^String make-str
  []
  "abc")
(defn ^String make-str
  ([] "abc")
  ([a] (str a "abc")))
; prefer
(defn make-str ^String [] "abc")
(defn make-str
  (^String [] "abc")
  (^String [a] (str a "abc")))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | false | 0.1.69 | 0.1.69 | 
when calls should have at least 1 expression after the condition.
; avoid
(when true)
(when (some-func))
; prefer
(when true (do-stuff))
(when (some-func) (do-stuff))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 1.2.0 | 
seq returns nil when given an empty collection. empty? is implemented as (not (seq coll)) so it's idiomatic to use seq directly.
; avoid
(not (empty? coll))
; prefer (chosen style :seq (default))
(seq coll)
; prefer (chosen style :not-empty)
(not-empty coll)
| Name | Default | Options | 
|---|---|---|
| :chosen-style | :seq | :seq,:not-empty | 
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 1.13 | 1.13 | 
NOTE: Requires Clojure version 1.12.0.
Uniform qualified method values are a new syntax for calling into java code. They must resolve to a single static or instance method and to help with that, a new metadata syntax can be used: ^[] aka ^{:param-tags []}. Types are specified with classes, each corrosponding to an argument in the target method: (^[long String] SomeClass/.someMethod 1 "Hello world!"). It compiles to a direct call without any reflection, guaranteeing optimal performance.
Given that, it is preferable to exclusively use method values.
; avoid
(.toUpperCase "noah")
(. "noah" toUpperCase)
; prefer
(^[] String/toUpperCase "noah")
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | false | 1.3.0 | 1.3.0 | 
In the ns form prefer :require :as over :require :refer over :require :refer :all. Prefer :require over :use; the latter form should be considered deprecated for new code.
; avoid
(ns examples.ns
  (:use clojure.zip))
; prefer
(ns examples.ns
  (:require [clojure.zip :as zip]))
(ns examples.ns
  (:require [clojure.zip :refer [lefts rights]]))
(ns examples.ns
  (:require [clojure.zip :refer :all]))
| Name | Default | Options | 
|---|---|---|
| :chosen-style | :as | :as,:refer,:all | 
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
A number of core functions take any number of arguments and return the arg if given only one. These calls are effectively no-ops, redundant, so they should be avoided.
Current list of clojure.core functions this linter checks:
->, ->>cond->, cond->>some->, some->>comp, partial, mergemin, max, distinct?; avoid
(-> x)
(->> x)
(cond-> x)
(cond->> x)
(some-> x)
(some->> x)
(comp x)
(partial x)
(merge x)
(min x)
(max x)
(distinct? x)
; prefer
x
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 1.16.0 | 1.16.0 | 
clojure.core/str calls .toString() on non-nil input. However, .toString() on a string literal returns itself, making it a no-op. Likewise, clojure.core/format unconditionally returns a string, making any calls to str on the results a no-op.
; avoid
(str "foo")
(str (format "foo-%s" some-var))
; prefer
"foo"
(format "foo-%s" some-var)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| false | true | false | 1.13 | 1.13 | 
NOTE: Requires Clojure version 1.12.0.
Uniform qualified method values are a new syntax for calling into java code. They must resolve to a single static or instance method and to help with that, a new metadata syntax can be used: ^[] aka ^{:param-tags []}. Types are specified with classes, each corrosponding to an argument in the target method: (^[long String] SomeClass/someMethod 1 "Hello world!")
If :param-tags is left off of a method value, then the compiler treats it as taking no arguments (a 0-arity static method or a 1-arity instance method with the instance being the first argument). And an _ can be used as a wild-card in the cases where there is only a single applicable method (no overloads).
These last two features are where there can be trouble. If, for whatever reason, the Java library adds an overload on type, then both the lack of :param-tags and a wild-card can lead to ambiguity. This is a rare occurence but risky/annoying enough that it's better to be explicit overall.
The styles are named after what they're looking for:
:missing checks that there exists a :param-tags on a method value.:wildcard checks that there are no usages of _ in an existing :param-tags.:both checks both conditions.; avoid (chosen style :both or :missing)
(java.io.File/mkdir (clojure.java.io/file "a"))
; avoid (chosen style :both or :wildcard)
(^[_ _] java.io.File/createTempFile "abc" "b")
; prefer (chosen style :both or :missing)
(^[] java.io.File/mkdir (clojure.java.io/file "a"))
; prefer (chosen style :both or :wildcard (default))
(^[String String] java.io.File/createTempFile "abc" "b")
| Name | Default | Options | 
|---|---|---|
| :chosen-style | :wildcard | :both,:missing,:wildcard | 
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 0.1 | 0.1 | 
repeatedly has an arity for limiting the number of repeats with take.
; avoid
(take 5 (repeatedly (range 10))
; prefer
(repeatedly 5 (range 10))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | false | false | 0.1 | 1.17.0 | 
Threading macros require more effort to understand so only use them with multiple args to help with readability.
Macros can be misinterpreted, leading to correct code being flagged:
(cond-> foo
  pred? (-> (assoc :hello 123)
            (dissoc :goodbye)))
; avoid
(-> x y)
(->> x y)
; prefer
(y x)
; avoid
(-> x (y z))
; prefer
(y x z)
; avoid
(->> x (y z))
; prefer
(y z x)
| Name | Default | Options | 
|---|---|---|
| :chosen-style | :inline | :inline,:avoid-collections | 
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | true | 1.0 | 1.0 | 
A macro that wraps a splicing unquote in a try-catch or try-finally can lead
to subtle hard to debug errors. Better to wrap the splicing unquote in a do
to force it into 'expression position'.
; avoid
`(try ~@body (finally :true))
; prefer
`(try (do ~@body) (finally :true))
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| true | true | false | 1.11 | 1.11 | 
Due to munging rules, underscores in namespaces can confuse tools and libraries which expect that underscores in class names should be dashes in Clojure.
; avoid
(ns foo_bar.baz_qux)
; prefer
(ns foo-bar.baz-qux)
| Enabled by default | Safe | Autocorrect | Version Added | Version Updated | 
|---|---|---|---|---|
| false | true | false | 1.8.0 | 1.8.0 | 
Because we can't (or won't) check for interop, *warn-on-reflection* should
be at the top of every file out of caution.
; avoid
(ns foo.bar)
(defn baz [a b] (+ a b))
; prefer
(ns foo.bar)
(set! *warn-on-reflection* true)
(defn baz [a b] (+ a b))
Can you improve this documentation?Edit on GitHub
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs | 
| ← | Move to previous article | 
| → | Move to next article | 
| Ctrl+/ | Jump to the search field |