type | layout | category | title |
---|---|---|---|
doc |
reference |
Basics |
Coding Conventions |
This page contains the current coding style for the Kotlin language.
In mixed-language projects, Kotlin source files should reside in the same source root as the Java source files, and follow the same directory structure (each file should be stored in the directory corresponding to each package statement).
In pure Kotlin projects, the recommended directory structure is to follow the package structure with the common root package omitted (e.g. if all the code in the project is in the "org.example.kotlin" package and its subpackages, files with the "org.example.kotlin" package should be placed directly under the source root, and files in "org.example.kotlin.foo.bar" should be in the "foo/bar" subdirectory of the source root.
Placing multiple declarations (classes, top-level functions or properties) in the same Kotlin source file is encouraged as long as these declarations are closely related to each other semantically.
If a Kotlin file contains a single class (potentially with related top-level declarations), its name should be the same
as the name of the class, with the .kt extension appended. If a file contains multiple classes, or only top-level declarations,
choose a name describing what the file contains, and name the file accordingly. Use camel humps with an uppercase first letter
(e.g. ProcessDeclarations.kt
).
The name of the file should describe what the code in the file does. Therefore, you should avoid using meaningless words such as "Util" in file names.
Kotlin follows the Java naming conventions. In particular:
- Names of packages are always lower case and do not use underscores (
org.example.myproject
). Using multi-word names is generally discouraged, but if you do need to use multiple words, simply concatenate them together. - Names of classes start with an upper case letter and use camel humps (
class DeclarationProcessor
) - Names of functions and properties start with a lower case letter and use camel humps and no underscores
(
fun processDeclarations()
,val declarationCount
) - Names of constants should use uppercase underscore-separated names (
const val MAX_COUNT = 8
) - For enum constants, it's OK to use either uppercase underscore-separated names
(
enum class Color { RED, GREEN }
) or regular camel-humps names starting with an uppercase letter, depending on the usage.
When using an acronym as part of a declaration name, capitalize it if it consists of two letters (IOStream
);
capitalize only the first letter if it is longer (HttpInputStream
).
The name of a class is usually a noun or a noun phrase explaining what the class is: List
, PersonReader
.
The name of a method is usually a verb or a verb phrase saying what the method does: close
, readPersons
.
The names should make it clear what the purpose of the entity is, so it's best to avoid using meaningless words
(Manager
, Wrapper
etc.) in names.
If a class has two properties which are conceptually the same but one is part of a public API and another is an implementation detail, use an underscore as the prefix for the name of the private property:
class C {
private val _elementList = mutableListOf<Element>()
val elementList: List<Element>
get() = _elementList
}
In tests (and only in tests), it's acceptable to use method names with spaces enclosed in backticks. (Note that such method names are currently not supported by the Android runtime.)
class MyTestCase {
@Test fun `ensure everything works`() {
}
}
In most cases, Kotlin follows the Java coding conventions.
Use 4 spaces for indentation. Do not use tabs.
Use K&R braces style: put the opening brace in the end of the line where the construct begins, and the closing brace on a separate line aligned vertically with the opening construct.
if (elements != null) {
for (element in elements) {
// ...
}
}
(Note: In Kotlin, semicolons are optional, and therefore line breaks are significant. The language design assumes K&R braces, and you may encounter surprising behavior if you try to use a different formatting style.)
Do omit semicolons whenever possible.
Put spaces around binary operators (a + b
). Exception: don't put spaces around the "range to" operator (0..i
).
Do not put spaces around unary operators (a++
)
Do put spaces between control flow keywords (if
, when
, for
and while
) and the corresponding opening parenthesis.
Do not put a space before an opening parenthesis in a primary constructor declaration, method declaration or method call.
class A(val x: Int)
fun foo(x: Int) { }
fun bar() {
foo(1)
}
Never put a space after (
, [
, or before ]
, )
.
Put a space after //
: // This is a comment
Do not put spaces around angle brackets used to specify type parameters: class Map<K, V> { ... }
Do not put spaces around ::
: Foo::class
, String::length
Do not put a space before ?
used to mark a nullable type: String?
Put a space before :
when it's used to separate a type and a supertype. Don't put a space before :
when it separates
a declaration and its type. Always put a space after :
.
interface Foo<out T : Any> : Bar {
fun foo(a: Int): T
}
Classes with a few arguments can be written in a single line:
class Person(id: Int, name: String)
Classes with longer headers should be formatted so that each primary constructor argument is in a separate line with indentation. Also, the closing parenthesis should be on a new line. If we use inheritance, then the superclass constructor call or list of implemented interfaces should be located on the same line as the parenthesis:
class Person(
id: Int,
name: String,
surname: String
) : Human(id, name) {
// ...
}
For multiple interfaces, the superclass constructor call should be located first and then each interface should be located in a different line:
class Person(
id: Int,
name: String,
surname: String
) : Human(id, name),
KotlinMaker {
// ...
}
Prefer putting a blank line after the class header to make it clear where the header ends and the class body begins.
Constructor parameters can use either the regular indent or the continuation indent (double the regular indent).
Generally, the contents of a class is sorted in the following order:
- Property declarations
- Secondary constructors
- Initializer blocks
- Method declarations
- Companion object
Do not sort the method declarations alphabetically or by visibility, and do not separate regular methods from extension methods. Instead, put related stuff together, so that someone reading the class from top to bottom would be able to follow the logic of what's happening. Choose an order (either higher-level stuff first, or vice versa) and stick to it.
Put nested classes next to the code that uses those classes. If the classes are intended to be used externally and aren't referenced inside the class, put them in the end, after the companion object.
When implementing an interface, keep the implementing members in the same order as members of the interface (if necessary, interspersed with additional private methods used for the implementation)
Always put overloads next to each other in a class.
If the function signature doesn't fit on a single line, use the following syntax:
fun longMethodName(
argument: ArgumentType = defaultValue,
argument2: AnotherArgumentType
): ReturnType {
// body
}
Prefer using an expression body for functions with the body consisting of a single expression.
fun foo(): Int { // bad
return 1
}
fun foo() = 1 // good
If the function has an expression body that doesn't fit in the same line as the declaration, put the =
sign on the first line.
Indent the expression body by 4 spaces.
fun f(x: String) =
x.length
If the expression body does not fit on one line, prefer using a block body instead.
If the body of a function is an if
, when
, try
or object
expression, the expression is not additionally indented;
the closing brace of the expression is on the same indentation level as where the function closing curly brace would be.
fun f(x: String) = when (x) {
"foo" -> ...
else -> ...
}
However, if the function header is long and can't be fit on the same line as function declaration (because the declaration is too long, the condition is complex, or the object expression needs a complex set of super-types), then the expression is on a new line and the closing brace of the expression is on the same indentation level as the head of the expression:
fun f(x: String, y: String) =
when (x.substring(y.length - 1) {
"foo" -> ...
else -> ...
}
If a function returns Unit, the return type should be omitted:
fun foo() { // ": Unit" is omitted here
}
Prefer declaring functions with default parameter values to declaring overloaded functions.
// Bad
fun foo() {
foo("a")
}
fun foo(a: String) { ... }
// Good
fun foo(a: String = "a") { ... }
For very simple read-only properties, consider one-line formatting:
val isEmpty: Boolean get() = size == 0
For more complex properties, always put get
and set
keywords on separate lines:
val foo: String
get() {
// ...
}
If you have a functional type or a type with type parameters which is used multiple times in a codebase, prefer defining a type alias for it:
typealias MouseClickHandler = (Any, MouseEvent) -> Unit
typealias PersonIndex = Map<String, Person>
Annotations are typically placed on separate lines, before the declaration to which they are attached, and with the same indentation:
@Target(AnnotationTarget.PROPERTY)
annotation class JsonExclude
Annotations without arguments may be placed on the same line:
@JsonExclude @JvmField
var x: String
A single annotation without arguments may be placed on the same line as the corresponding declaration:
@Test fun foo() { ... }
File annotations are placed after the file comment (if any), before the package
statement, and are separated from package
with a blank line (to emphasize the fact that they target the file and not the package).
/** License, copyright and whatever */
@file:JvmName("FooBar")
package foo.bar
In lambda expressions, spaces should be used around the curly braces, as well as around the arrow which separates the parameters from the body. If a call takes a single lambda, it should be passed outside of parentheses whenever possible.
list.filter { it > 10 }.map { element -> element * 2 }
In lambdas which are short and not nested, it's recommended to use the it
convention instead of declaring the parameter
explicitly. In nested lambdas with parameters, parameters should be always declared explicitly.
When declaring parameter names in a multiline lambda, put the names on the first line, followed by the arrow and the newline:
appendCommaSeparated(properties) { prop ->
val propertyValue = prop.get(obj) // ...
}
Avoid using multiple labeled returns in a lambda. Consider restructuring the lambda so that it will have a single exit point. If that's not possible or not clear enough, convert the lambda into an anonymous function.
Do not use a labeled return for the last statement in a lambda.
Use the named argument syntax when a method takes multiple parameters of the same primitive type, or for parameters of Boolean
type,
unless the meaning of all parameters is absolutely clear from context.
drawSquare(x = 10, y = 10, width = 100, height = 100, fill = true)
Do put spaces around the =
sign separating the argument name and value.
In long argument lists, put a line break after the opening parenthesis. Group multiple closely related arguments on the same line.
drawSquare(
x = 10, y = 10,
width = 100, height = 100,
fill = true
)
When wrapping chained calls, put the . character or the ?.
operator on the next line, with a single indent:
val anchor = owner
.firstChild!!
.siblings(forward = true)
.dropWhile { it is PsiComment || it is PsiWhiteSpace }
The first call in the chain usually should have a line break before it, but it's OK to omit it the code makes more sense that way.
Prefer using the expression form of try
, if
and when
. Example:
return if (x) foo() else bar()
The above is preferable to:
if (x)
return foo()
else
return bar()
Prefer using if
for binary conditions instead of when
. Instead of
when (x) {
null -> ...
else -> ...
}
use if (x == null) ... else ...
Prefer using when
if there are three or more options.
In a when
statement, if a branch is more than a single line, always separate it from adjacent case blocks with a blank line:
private fun parsePropertyValue(propName: String, token: Token) {
when (token) {
is Token.ValueToken ->
callback.visitValue(propName, token.value)
Token.LBRACE -> { // ...
}
Put short branches on the same line as the condition, without braces.
when (foo) {
true -> bar() //good
false -> { baz() } //bad
}
If you need to use a nullable Boolean
in a conditional statement, use if (value == true)
or if (value == false)
checks.
Prefer using higher-order functions (filter
, map
etc.) to loops. Exception: forEach
(prefer using a regular for loop instead,
unless the receiver of forEach
is nullable or forEach
is used as part of a longer call chain).
When making a choice between a complex expression using multiple higher-order functions and a loop, understand the cost of the operations being performed in each case and keep performance considerations in mind.
Use the until
function to loop over an open range:
for (i in 0..n - 1) { ... } // bad
for (i in 0 until n) { ... } // bad
In some cases functions with no arguments might be interchangeable with read-only properties. Although the semantics are similar, there are some stylistic conventions on when to prefer one to another.
Prefer a property over a function when the underlying algorithm:
- does not throw
- has a
O(1)
complexity - is cheap to calculate (or caсhed on the first run)
- returns the same result over invocations
Use extension functions liberally. Every time you have a function that works primarily on an object, consider making it an extension function accepting that object as a receiver. To minimize API pollution, restrict the visibility of extension functions as much as it makes sense. As necessary, use local extension functions, member extension functions, or top-level extension functions with private visibility.
Declare a function as infix only when it works on two objects which play a similar role. Good examples: and
, to
, zip
.
Bad example: add
.
Don't declare a method as infix if it mutates the receiver object.
If you declare a factory function for a class, don't give it the same name as the class itself. Use a distinct name making it clear why the behavior of the factory function is special. Example:
class Point(val x: Double, val y: Double) {
companion object {
fun fromPolar(angle: Double, radius: Double) = Point(...)
}
}
If you have an object with multiple overloaded constructors that don't call different superclass constructors and can't be reduced to a single constructor with default argument values, prefer to replace the overloaded constructors with factory functions.
If a declaration has multiple modifiers, always put them in the following order:
public / protected / private / internal
final / open / abstract / sealed / const
external
override
lateinit
tailrec
vararg
suspend
inner
enum / annotation
companion
inline
infix
operator
data
A public function/method returning an expression of a platform type must declare its kotlin type explicitly:
fun apiCall(): String = MyJavaApi.getProperty("name")
Any property (package-level or class-level) initialised with an expression of a platform type must declare its kotlin type explicitly:
class Person {
val name: String = MyJavaApi.getProperty("name")
}
A local value initialised with an expression of a platform type may or may not have a type declaration:
fun main(args: Array<String>) {
val name = MyJavaApi.getProperty("name")
println(name)
}
Use apply
for initialization:
val foo = createBar().apply {
property = value
init()
}
Use also
over apply
if the receiver is used for anything other than setting properties or function calls on it,
or if the receiver is not used at all in the lambda.
class Baz {
var currentBar: Bar?
val observable: Observable
val foo = createBar().also {
currentBar = it
observable.registerCallback(it)
}
}
Prefer also
over apply
if there already are multiple receivers in scope, especially if you make calls on any outer receivers:
class Foo {
fun Bar.baz() {
val stuff = callSomething().also {
it.init()
this@baz.registerCallback(it)
}
}
}
Prefer apply
/run
over with
if the receiver is nullable.
getNullable()?.run {
init()
}
getNullable()?.apply {
init()
}
Prefer run
/with
over apply
if the returned value is not used
view.run {
textView.text = "Hello World"
progressBar.init()
}
with(view) {
textView.text = "Hello World"
progressBar.init()
}
Choose one of the above and use it consistently.
Prefer let
over run
in method chains that transform the receiver
val baz: Baz = foo.let { createBar(it) }.convertBarToBaz()
// or with function references
val baz: Baz = foo.let(::createBar).convertBarToBaz()
Prefer using string templates to string concatenation.
Don't use curly braces when inserting a simple variable into a string template. Use curly braces only for longer expressions.
println("$name has ${children.size} children")
Prefer to use multiline strings instead of embedding \n
escape sequences into regular string literals.
Use trimMargin
to maintain indentation in multiline strings:
assertEquals("""Foo
|Bar""".trimMargin(), value)
When writing libraries, it's recommended to follow an additional set of rules to ensure API stability:
- Always explicitly specify member visibility (to avoid accidentally exposing declarations as public API)
- Always explicitly specify function return types and property types (to avoid accidentally changing the return type when the implementation changes)
- Provide KDoc comments for all public methods (to support generating documentation for the library)