Configuration
If you have very specific preferences for configuring your formatting, please make sure you have read this entire document, and identified every single parameter available, along with their interactions.
Many parameters do not have ultimate authority and might be overridden by (possibly, default) values of other parameters.
Configuration for scalafmt is defined in a plain text file .scalafmt.conf
using HOCON syntax.
Here is an example .scalafmt.conf:
align.preset = more // For pretty alignment.
maxColumn = 1234
🚧 Before using specific configuration make sure that your project agrees on the standards. Settings such as
maxColumncould be a source of issues if different tools such as an IDE uses a different default value.
Most popular
maxColumn
Default: maxColumn = 80
- Keep in mind that 80 characters fit perfectly on a split laptop screen with regular resolution.
- GitHub mobile view only shows 80 characters and sometimes you might review code on your phone.
- Consider refactoring your code before choosing a value above 100.
assumeStandardLibraryStripMargin
This parameter simply says the .stripMargin method was not redefined by the
user to assign special meaning to indentation preceding the | character.
Hence, that indentation can be modified.
# Defaults
assumeStandardLibraryStripMargin = false
align.stripMargin = true
If true, lines starting with the margin character | (or another if specified
in the .stripMargin(...) call) will be indented differently.
If align.stripMargin is true, they will align with the opening triple-quote
""" in interpolated and raw string literals. Otherwise, they will be indented
relative to the start of the opening line.
val example1 =
s"""Examples:
| * one
| * two
| * $three
|""".stripMargin
val example1 =
s"""Examples:
| * one
| * two
| * $three
|""".stripMargin
Config for this example:
assumeStandardLibraryStripMargin = true
align.stripMargin = true
val example1 =
s"""|Examples:
| * one
| * two
| * $three
|""".stripMargin
val example1 =
s"""|Examples:
| * one
| * two
| * $three
|""".stripMargin
Config for this example:
assumeStandardLibraryStripMargin = true
align.stripMargin = false
The pipe character can immediately follow the opening """
val example2 =
s"""|Examples:
| * one
| * two
| * $three
|""".stripMargin
val example2 =
s"""|Examples:
| * one
| * two
| * $three
|""".stripMargin
Config for this example:
assumeStandardLibraryStripMargin = true
align.stripMargin = true
Version
The version parameter specifies the release of the formatter to be used.
If the version requested is different from the version of the installed
formatter, the correct release will be downloaded dynamically.
Since v3.1.0, the
versionparameter is required to be specified explicitly.
Since that requires access to artifact repositories, please see more on that on the Installation page.
Scala Dialects
The formatter supports various dialects defined and supported by scalameta parser. However,
for any given file, one and only one dialect is supported, and runner.dialect must be
used to select it.
Since v3.1.0, the
runner.dialectparameter is required to be specified explicitly.
Available dialects are:
- Scala 2
scala211scala212scala212source3scala213scala213source3
- Scala 3
scala3: most recent releaseScala3Future: scala3 plus experimental, unreleased features- specific versions from
Scala30toScala36
- sbt-specific
sbt0137sbt1
You can also specify runner.dialect for a subset of files using fileOverride:
fileOverride {
"glob:**/scala3-subproject/src/main/scala/**" {
runner.dialect = scala3
}
}
runner.dialectOverride
Since v3.6.0
Using this section, you can explicitly set
scalameta Dialect fields
(directly or via their associated withXxxYyy methods).
// top-level def: unsupported by scala213
def foo = // significant indent: unsupported by scala213
bar
baz
end foo // end marker: unsupported by scala213
// top-level def: unsupported by scala213
def foo = // significant indent: unsupported by scala213
bar
baz
end foo // end marker: unsupported by scala213
Config for this example:
runner.dialect = scala213
runner.dialectOverride.withAllowToplevelTerms = true
runner.dialectOverride.withAllowEndMarker = true
runner.dialectOverride.allowSignificantIndentation = true
Scala 3
Since v3.0.0, scalafmt supports Scala 3 features that can be enabled by changing
the dialect of the scalameta parser.
runner.dialect = scala3
---
open class MyOpenClass()
val myQuote = '{ expr }
val mySplice = ${ expr }
enum class Weekday {
case Monday, Tuesday, Wednesday, Thursday, Friday
}
Please also see rewrite rules for Scala 3.
Scala 2 with -Xsource:3
Also since v3.0.0, if using -Xsource:3 option for Scala 2 compiler, you can change the
dialect to Scala213Source3, which will allow to format some of the new syntax backported
from Scala 3.
Presets
Some sections provide preset values to set multiple parameters at once. These
are always accessed via the preset key of the appropriate section, including
top-level.
Top-level presets
preset=default
This preset is implicit and sets all values to their defaults.
preset=IntelliJ
This preset is defined as
preset = default
indent.defnSite = 2
optIn.configStyleArguments = false
preset=defaultWithAlign
This preset is defined as
preset = default
align.preset = more
preset=Scala.js
This preset intends to approximate the
style used for scala.js.
It uses modified detection of config-style formatting:
- according to Sébastien Doeraene, config-style should be driven solely by presence of a dangling closing parenthesis
- to achieve that, use a combination of
danglingParentheses.xxxSite = falseandnewlines.configStyle.xxxSite.prefer = true
The preset itself is defined as:
preset = default
binPack.preset = true
align {
openParenCtrlSite = false
arrowEnumeratorGenerator = false
tokens = [ caseArrow ]
}
danglingParentheses {
callSite = false
defnSite = false
}
docstrings.style = Asterisk
indent.callSite = 4
newlines {
avoidInResultType = true
neverBeforeJsNative = true
sometimesBeforeColonInMethodReturnType = false
}
runner.optimizer {
callSite {
minSpan = 500
minCount = 5
}
defnSite {
minSpan = 500
minCount = 5
}
}
Appending to preset collections
If, instead of redefining the default or preset value of a list or a map parameter, you'd like to append to it, use the following syntax:
// set
a.b.c.list = [ ... ]
a.b.c.dict = { ... }
// append; "+" must be the only key
a.b.c.list."+" = [ ... ]
a.b.c.dict."+" = { ... }
Indentation
indent.main
Since v3.0.0.
This parameter controls the primary code indentation. Various context-specific overrides are defined below, within this section.
Default: indent.main = 2
indent.significant
Since v3.0.0.
This parameter controls the amount of significant indentation used when optional braces rules apply.
By default, equals to indent.main.
object a {
if (foo)
bar
else
baz
if foo then
bar
bar
else
baz
baz
}
object a {
if (foo)
bar
else
baz
if foo then
bar
bar
else
baz
baz
}
Config for this example:
runner.dialect = scala3
indent.main = 2
indent.significant = 3
indent.callSite
This parameter controls indentation of method call arguments.
Default: indent.callSite = 2
Example:
function(
argument1, // indented by 2
""
)
function(
argument1, // indented by 2
""
)
Config for this example:
indent.callSite = 2
indent.binPackCallSite
This parameter controls indentation of method call arguments when they are
binpacked (that is, spanning multiple lines but not formatted using config
style). Defaults to indent.callSite.
Since v3.8.3.
Example:
func(1, // next indented by 4
"")
func( // next indented by 2
1, "")
func(1, // next indented by 4
"")
func( // next indented by 2
1,
"")
Config for this example:
newlines.configStyle.callSite.prefer = false
danglingParentheses.callSite = false
binPack.callSite = always
indent.callSite = 2
indent.binPackCallSite = 4
indent.ctrlSite
This parameter controls indentation within control expressions (if/while/etc).
If not set, the value of indent.callSite applies.
Since v3.0.0.
if (
foo && // indented by 4
bar
) {
baz
}
if (
foo && // indented by 4
bar
) {
baz
}
Config for this example:
indent.callSite = 2
indent.ctrlSite = 4
if (
foo && // indented by 2
bar
) {
baz
}
if (
foo && // indented by 2
bar
) {
baz
}
Config for this example:
indent.callSite = 2
indent.defnSite
Default: indent.defnSite = 4
Same as indent.callSite except for definition site. Example:
def function(
parameter1: Type1 // indented by 4
): ReturnType
def function(
parameter1: Type1 // indented by 4
): ReturnType
Config for this example:
indent.defnSite = 4
indent.binPackDefnSite
This parameter controls indentation of method parameters when they are
binpacked (that is, spanning multiple lines but not formatted using config
style). Defaults to indent.defnSite.
Since v3.8.3.
Example:
def func(a: Int, // next indented by 4
b: String)
def func( // next indented by 2
a: Int, b: String)
def func(a: Int, // next indented by 4
b: String)
def func( // next indented by 2
a: Int,
b: String)
Config for this example:
newlines.configStyle.defnSite.prefer = false
danglingParentheses.defnSite = false
binPack.defnSite = always
indent.defnSite = 2
indent.binPackDefnSite = 4
indent.ctorSite
Since v2.5.0.
Applies to constructors. Defaults to indent.defnSite.
class A(
field1: Type1 // indented by 4
) {
def function2(
parameter1: Type1 // indented by 2
): ReturnType = None
}
class A(
field1: Type1 // indented by 4
) {
def function2(
parameter1: Type1 // indented by 2
): ReturnType = None
}
Config for this example:
indent.ctorSite = 4
indent.defnSite = 2
indent.matchSite
Since v3.4.4.
If set, applies custom indentation to case clauses in match expressions.
Default: indent.matchSite = null
object a:
x match
case _: Aaaaaa |
_: Bbbbbb |
_: Cccccc =>
end match
object a:
x match
case _: Aaaaaa |
_: Bbbbbb |
_: Cccccc =>
end match
Config for this example:
maxColumn = 20
indent.matchSite = 0
runner.dialect = scala3
indent.caseSite
Since v3.0.0.
Applies indentation to patterns in case clauses.
Default: indent.caseSite = 4
x match {
case _: Aaaaaa |
_: Bbbbbb |
_: Cccccc =>
}
x match {
case _: Aaaaaa |
_: Bbbbbb |
_: Cccccc =>
}
Config for this example:
maxColumn = 20
indent.caseSite = 5
indent.extendSite
Default: indent.extendSite = 4
This parameter defines indentation used for the extends A with B or derives A, B sequences
in a template (class, trait, object, enum, etc.).
trait Foo
extends A {
def foo: Boolean =
true
}
trait Foo extends A {
def foo: Boolean = true
}
Config for this example:
indent.extendSite = 4
maxColumn = 20
indent.withSiteRelativeToExtends
Since v2.5.0.
This parameter defines additional indentation used for the with elements of an
extends A with B sequence in a template.
Default: indent.withSiteRelativeToExtends = 0
trait Foo
extends A
with B
with C
with D
with E {
def foo: Boolean = true
}
trait Foo extends A with B with C with D with E {
def foo: Boolean = true
}
Config for this example:
indent.extendSite = 4
indent.withSiteRelativeToExtends = 2
maxColumn = 30
indent.commaSiteRelativeToExtends
Since v3.0.0
This parameter defines additional indentation used for the post-comma elements
of an extends A, B or derives A, B sequences of a template.
Added to support Scala 3, which allows to specify multiple parents with a comma.
Default: indent.commaSiteRelativeToExtends = 2
trait Foo
extends A,
B,
C,
D,
E {
def foo: Boolean =
true
}
trait Foo extends A, B, C, D, E {
def foo: Boolean = true
}
Config for this example:
runner.dialect = scala3
indent.extendSite = 4
indent.commaSiteRelativeToExtends = 4
maxColumn = 20
indent.extraBeforeOpenParenDefnSite
Since v3.0.0
This parameter applies to definitions and sets extra indentation (relative to the indentation of the body) used for parameter groups when
- the definition has a body (that needs differentiating from)
- newlines.beforeOpenParenDefnSite is set
Default: indent.extraBeforeOpenParenDefnSite = 0
case class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
abstract class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) {
def fooDef
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
}
def fooDef(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) = {
// body
}
case class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
abstract class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) {
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
}
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) = {
// body
}
Config for this example:
maxColumn = 25
newlines.beforeOpenParenDefnSite = fold
case class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
abstract class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) {
def fooDef
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
}
def fooDef(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) = {
// body
}
case class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
abstract class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) {
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
}
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) = {
// body
}
Config for this example:
maxColumn = 25
indent.extraBeforeOpenParenDefnSite = 2
newlines.beforeOpenParenDefnSite = fold
case class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
abstract class fooClass
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) {
def fooDef
(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String)
}
def fooDef(foo1: String)
(
foo2: String,
foo3: String
)(foo5: String) = {
// body
}
case class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
abstract class fooClass
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) {
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String)
}
def fooDef
(foo1: String)
(foo2: String, foo3: String)
(foo5: String) = {
// body
}
Config for this example:
maxColumn = 25
indent.extraBeforeOpenParenDefnSite = -1
newlines.beforeOpenParenDefnSite = fold
indent.relativeToLhsLastLine
When the left-hand side of an infix or match expression is itself broken over
several lines, with the last line indented relative to the first line, this flag
determines whether the indent is relative to the first or the last line.
This parameter takes a list of values including:
match: applies to match expressionsinfix: applies to infix expressions
Default: indent.relativeToLhsLastLine = []
foo // c1
.bar match {
case baz => qux
}
foo // c1
.bar infix { case baz =>
qux
}
foo // c1
.bar match {
case baz => qux
}
foo // c1
.bar infix {
case baz => qux
}
Config for this example:
indent.relativeToLhsLastLine = []
foo // c1
.bar match {
case baz => qux
}
foo // c1
.bar infix { case baz =>
qux
}
foo // c1
.bar match {
case baz => qux
}
foo // c1
.bar infix {
case baz => qux
}
Config for this example:
indent.relativeToLhsLastLine = [match, infix]
indent.fewerBraces
This parameter controls whether extra indent.main is added
to the sole argument of a method call using the "fewer braces"
syntax. The following values are supported:
always: always applies extra indentnever: doesn't apply any extra indentbeforeSelect: applies extra indent only to fewer-braces expressions followed by a.select
Default: indent.fewerBraces = "never"
In Scala 3.3.0, only
neverprovides compiler-compatible code. Other options will work in 3.3.1-RC1 and later (see Parser section). Also,neveris implicitly forced if indentation width is less than 2.
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
Config for this example:
runner.dialect = Scala3
indent.significant = 3
indent.fewerBraces = always
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
Config for this example:
runner.dialect = Scala3
indent.significant = 3
indent.fewerBraces = never
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
bar:
2 + 2
foo.bar:
2 + 2
foo:
2 + 2
.bar:
3 + 3
.baz // c
.qux
Config for this example:
runner.dialect = Scala3
indent.significant = 3
indent.fewerBraces = beforeSelect
indent.afterInfixSite
This parameter controls indentation after an infix operator (not including
assignments). Defaults to indent.main.
Since v3.8.3.
Indent for binPack.callSite
Normally, even when binpacking, there's a new level of indentation added for each opening parenthesis starting a nested argument clause (regardless whether the first argument follows the opening parenthesis on the same line or on a separate one); the parameters below modify this behaviour.
See also indent.binPackCallSite.
binPack.indentCallSiteOnce
When this parameter is enabled, only one level is added to the outermost call, regardless of the number of nested parentheses.
Default: binPack.indentCallSiteOnce = false
With the parameter enabled:
foo(bar1(baz1(qux1,
qux2), baz2),
bar2(baz3, baz4))
foo(bar1(baz1(qux1, qux2), baz2), bar2(baz3, baz4))
Config for this example:
binPack.callSite = true
binPack.indentCallSiteOnce = true
indent.callSite = 2
maxColumn = 20
With the parameter disabled:
foo(bar1(baz1(qux1,
qux2), baz2),
bar2(baz3, baz4))
foo(bar1(baz1(qux1, qux2), baz2), bar2(baz3, baz4))
Config for this example:
binPack.callSite = true
binPack.indentCallSiteOnce = false
indent.callSite = 2
maxColumn = 20
binPack.indentCallSiteSingleArg
When this parameter is disabled, no indentation is added for same-line single-arg cases; the assumption is that if the argument expression spans multiple lines, it will introduce its own indentation.
Default: binPack.indentCallSiteSingleArg = true
With the parameter enabled:
foo(bar(baz.qux(
xyz + zyx
)))
foo(
bar((_, _) =>
baz { qux =>
noop
} baz { qux =>
noop
} baz //
{ qux =>
noop
} baz { qux =>
noop
}
)
)
foo(bar(baz.qux(xyz + zyx)))
foo(bar((_, _) =>
baz { qux =>
noop
} baz { qux =>
noop
} baz //
{ qux =>
noop
} baz { qux =>
noop
}))
Config for this example:
binPack.callSite = true
binPack.indentCallSiteSingleArg = true
indent.callSite = 2
maxColumn = 20
With the parameter disabled:
foo(bar(
baz.qux(xyz + zyx)
))
foo(
bar((_, _) =>
baz { qux =>
noop
} baz { qux =>
noop
} baz //
{ qux =>
noop
} baz { qux =>
noop
}
)
)
foo(bar(baz.qux(xyz + zyx)))
foo(bar((_, _) =>
baz { qux =>
noop
} baz { qux =>
noop
} baz //
{ qux =>
noop
} baz { qux =>
noop
}))
Config for this example:
binPack.callSite = true
binPack.indentCallSiteSingleArg = false
indent.callSite = 2
maxColumn = 20
indent.infix
Prior to v3.8.4, this section was called
indentOperator.
Normally, the first eligible break inside a chain of infix operators is indented by 2.
This group of parameters allows overriding which infix operators, and in which context, are eligible to be exempted from this, with indentation omitted.
If you wish to disable this functionality, set
indent.infix.excludeRegex = '^$'.
indent.infix.exemptScope
Added in 3.4.0, this parameter determines when an infix operator can be exempted from applying continuation indentation.
It accepts the following values, to determine the context in which infix operators are eligible to be exempted from the default indentation rule:
oldTopLevel(default): "top-level" infix operators- the definition of top-level historically refers to infix expressions whose direct parent is
a block (typically as the last statement in that block),
if/while(as condition or body), or case clause (as pattern or body) - this value replaced deprecated
indentOperator.topLevelOnly=true - this approach is also somewhat inconsistent with what it was intended to accomplish, and kept only for backwards compatibility; please consider using one of the alternatives
- the definition of top-level historically refers to infix expressions whose direct parent is
a block (typically as the last statement in that block),
aloneEnclosed: infix operators which are enclosed in braces or parens as the only statement in a block or body of a braces-enclosed lambda function; anif/whilecondition; the only argument of a method call; or similar;- it also includes a few scenarios where parens can be omitted, such case clause patterns,
conditions in new scala3
if-thenandwhile-dosyntax, etc. - however, block braces are not optional
- it also includes a few scenarios where parens can be omitted, such case clause patterns,
conditions in new scala3
aloneArgOrBody: infix operators as anif/whilecondition; an argument of a method call; the only statement in a block; entire body of an assignment, case clause, control statement, etc;- it is intended to help implement a requirement of the scala-js coding style.
all: all infix operators- this value replaced deprecated
indentOperator.topLevelOnly=false
- this value replaced deprecated
notAssign(since v3.8.4): any non-assignment operator- this value expanded upon deprecated
verticalAlignMultilineOperatorswhich now simply maps to{ exemptScope = notAssign, excludeRegex = ".*" }
- this value expanded upon deprecated
notWithinAssign(since v3.8.4): any infix not part of a larger assignment expression
function(
a &&
b,
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
function(
a &&
b,
a &&
b
)
function(a &&
b)(a &&
b)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
Config for this example:
indent.infix.exemptScope = oldTopLevel
function(
a &&
b,
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
function(
a &&
b,
a &&
b
)
function(a &&
b)(a &&
b)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
Config for this example:
indent.infix.exemptScope = all
function(
a &&
b,
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
function(
a &&
b,
a &&
b
)
function(a &&
b)(a &&
b)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
Config for this example:
indent.infix.exemptScope = aloneEnclosed
function(
a &&
b,
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
function(
a &&
b,
a &&
b
)
function(a &&
b)(a &&
b)
function(
a &&
b
)(
a &&
b
)
function {
a &&
b
}
Config for this example:
indent.infix.exemptScope = aloneArgOrBody
indent.infix.excludeRegex
Defines a regular expression for excluded infix operators. If an eligible operator matches, it will not be indented.
In v3.1.0, this parameter was renamed from indentOperator.exclude.
Default: indent.infix.excludeRegex = "^(&&|\\|\\|)$"
indent.infix.includeRegex
Defines a regular expression for included infix operators. If an eligible operator matches and is not excluded explicitly by indent.infix.excludeRegex, it be will indented.
In v3.1.0, due to conflict with built-in HOCON keyword, this parameter was
renamed from indentOperator.include.
Default: indent.infix.includeRegex = ".*"
indent.infix.preset
default- use defaults for all fields
spray(alsoakka)- set
indent.infix.excludeRegex = "^$"andindent.infix.includeRegex = "^.*=$"
- set
Alignment
Alignment describes formatting which inserts additional spaces to align certain tokens on different lines vertically.
Apart from a few special cases, the way alignment works is as follows:
- for each line, alignment stops are identified, by looking up each token in
align.tokens(matching token, owner and, if specified, the owner's parent) - for two candidate lines, respective alignment stops are compared (first stop on one line to the first one on the other, etc); the only exception are the single-line comments which are compared regardless of their stop position
- two tokens will match if:
- both tokens have the same token category; a token's category is the value
associated with its type in
align.tokenCategorymapping or, if missing, its type - both owners have the same tree category; similarly, a tree's category is the
value for its type in
align.treeCategorymapping or the type itself - both owners belong to the same "statement container"; this is determined internally and usually selects the nearest containing block, template, match, argument or parameter group.
- both tokens have the same token category; a token's category is the value
associated with its type in
- for each token that has matches in the surrounding lines:
- we'll determine the amount of extra space needed to be added before that token, to align it on the right with matching tokens
- however, if there was no space before the token, and
align.delayUntilSpaceis set, that extra space will be added to the next space on its line, thus aligning subsequent token on the left.
Align has several nested fields, which you can customize. However, it comes with four possible presets: none, some, more, & most.
align.preset
Default: some
align.preset=none
x match { // false for case arrows
case 2 => 22 // also comments!
case 22 => 222 // don't align me!
}
x match { // false for case arrows
case 2 => 22 // also comments!
case 22 => 222 // don't align me!
}
Config for this example:
align.preset = none
Pro tip: Enable this setting to minimize git diffs/conflicts from renamings and other refactorings, without having to ignore whitespace changes in diffs or use
--ignore-all-spaceto avoid conflicts when git merging or rebasing.
Starting with the introduction of
align.stripMarginparameter in v2.5.0, one must explicitly enable it to get earlier behaviour ofalign.preset=none. See assumeStandardLibraryStripMargin.
align.preset=some
x match { // true for case arrows
case 2 => 22
case 22 => 222
}
val x = 2 // false for assignment
val xx = 22
case object B
extends A // false for `extends`
case object BB extends A
x match { // true for case arrows
case 2 => 22
case 22 => 222
}
val x = 2 // false for assignment
val xx = 22
case object B extends A // false for `extends`
case object BB extends A
Config for this example:
align.preset = some
align.preset=more
val x = 2 // true for assignment
val xx = 22
case object B extends A // true for `extends`
case object BB extends A
q -> 22 // true for various infix operators
qq -> 3 // and also comments!
for {
x <- List(1) // true for alignment enumerator
yyy <- List(2)
} yield x ** yyy
x match { // true for multiple tokens across multiple lines
case 1 => 1 -> 2 // first
case 11 => 11 -> 22 // second
// A blank line separates alignment blocks.
case `ignoreMe` => 111 -> 222
}
// Align assignments of similar type.
def name = column[String]("name")
def status = column[Int]("status")
val x = 1
val xx = 22
// Align sbt module IDs.
libraryDependencies ++= Seq(
"org.scala-lang" % "scala-compiler" % scalaVersion.value,
"com.lihaoyi" %% "sourcecode" % "0.1.1"
)
val x = 2 // true for assignment
val xx = 22
case object B extends A // true for `extends`
case object BB extends A
q -> 22 // true for various infix operators
qq -> 3 // and also comments!
for {
x <- List(1) // true for alignment enumerator
yyy <- List(2)
} yield x ** yyy
x match { // true for multiple tokens across multiple lines
case 1 => 1 -> 2 // first
case 11 => 11 -> 22 // second
// A blank line separates alignment blocks.
case `ignoreMe` => 111 -> 222
}
// Align assignments of similar type.
def name = column[String]("name")
def status = column[Int]("status")
val x = 1
val xx = 22
// Align sbt module IDs.
libraryDependencies ++= Seq(
"org.scala-lang" % "scala-compiler" % scalaVersion.value,
"com.lihaoyi" %% "sourcecode" % "0.1.1"
)
Config for this example:
maxColumn = 80
align.preset = more
align.preset=most
for {
// align <- with =
x <- List()
yyy = 2
// aligns body by arrow
zzz <- new Integer {
def value = 3
}
} yield x
for {
// align <- with =
x <- List()
yyy = 2
// aligns body by arrow
zzz <- new Integer {
def value = 3
}
} yield x
Config for this example:
align.preset = most
Note. Only for the truest vertical aligners.
align.tokens
Default: [caseArrow]
An align token contains a code (the string literal of an operator of token) and a list of
owners; each owner entry in turn contains an optional regex (the kind of the closest tree
node that owns that token), and a list of parents (to match the tree containing the owner of
the token).
To find the
ownerpart for a custom tree, look for its type prefix using Scastie Playground or AST Explorer.
The special code
//is used for single-line comments. Also, since v3.3.1, this includes multi-line comments/* ... */which do not themselves contain newlines but are followed by one (i.e., can trivially be changed to a//comment).
x match {
case 1 => 1 -> 2
case 11 => 11 -> 22
}
x match {
case 1 => 1 -> 2
case 11 => 11 -> 22
}
Config for this example:
align.tokens = [{
code = "=>"
owners = [{
regex = "Case"
}]
}]
val x = List(
"org.scala-lang" %% "scala-compiler" % scalaVersion.value,
"com.lihaoyi" %% "sourcecode" % "0.1.1"
)
val x = List(
"org.scala-lang" %% "scala-compiler" % scalaVersion.value,
"com.lihaoyi" %% "sourcecode" % "0.1.1"
)
Config for this example:
align.tokens = [
{
code = "%"
owners = [{
regex = "Term.ApplyInfix"
}]
}, {
code = "%%"
owners = [{
regex = "Term.ApplyInfix"
}]
}
]
case class Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) {
def Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) = ???
}
case class Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) {
def Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) = ???
}
Config for this example:
align.tokens."+" = [{
code = ":"
owners = [{
regex = "Term\\.Param"
parents = [ "Ctor\\.Primary" ]
}]
}]
case class Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) {
def Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) = ???
}
case class Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) {
def Foo(
firstParam: Int,
secondParam: String,
thirdParam: Boolean
) = ???
}
Config for this example:
align.tokens."+" = [{
code = ":"
owners = [{
parents = [ "Defn\\." ]
}]
}]
align.arrowEnumeratorGenerator
Default: align.arrowEnumeratorGenerator = false
for {
x <- new Integer {
def value = 2
}
} yield x
for {
x <- new Integer {
def value = 2
}
} yield x
Config for this example:
align.arrowEnumeratorGenerator = false
for {
x <- new Integer {
def value = 2
}
} yield x
for {
x <- new Integer {
def value = 2
}
} yield x
Config for this example:
align.arrowEnumeratorGenerator = true
align.closeParenSite
This parameter controls whether to align the closing parentheses when we
aligned the opening one (see openParenXxxSite parameters below) and the
respective danglingParentheses.xxxSite is set.
Since v3.0.0.
Default: align.closeParenSite = false
function(arg1, // align by (
arg2,
arg3
)
function(arg1, // align by (
arg2,
arg3)
Config for this example:
align.closeParenSite = true
align.openParenCallSite = true
danglingParentheses.callSite = true
function(arg1, // align by (
arg2,
arg3
)
function(arg1, // align by (
arg2,
arg3)
Config for this example:
align.closeParenSite = false
align.openParenCallSite = true
danglingParentheses.callSite = true
align.openParenCallSite
Default: align.openParenCallSite = false
Default changed from
truetofalsein v1.6.
foo(arg1, arg2)
function(arg1, // align by (
arg2,
arg3
)
function(argument1, argument2)
foo(arg1, arg2)
function(arg1, // align by (
arg2,
arg3)
function(
argument1,
argument2)
Config for this example:
align.openParenCallSite = true
foo(arg1, arg2)
function(
arg1, // no align by (
arg2,
arg3
)
function(argument1, argument2)
foo(arg1, arg2)
function(arg1, // no align by (
arg2,
arg3)
function(
argument1,
argument2)
Config for this example:
align.openParenCallSite = false
align.openBracketCallSite
Since v3.0.4.
If set explicitly, will be used for the left bracket in type arguments,
instead of align.openParenCallSite.
align.openParenCtrlSite
This parameter controls alignment after ( in if/while/for.
Default: align.openParenCtrlSite = false
align.openParenDefnSite
Default: align.openParenDefnSite = false
Default changed from
truetofalsein v1.6.
class IntString(int: Int,
string: String
)
class IntStringLong(int: Int,
string: String,
long: Long
)
class IntString(int: Int, string: String)
class IntStringLong(int: Int,
string: String,
long: Long)
Config for this example:
align.openParenDefnSite = true
class IntString(
int: Int,
string: String
)
class IntStringLong(
int: Int,
string: String,
long: Long
)
class IntString(int: Int, string: String)
class IntStringLong(
int: Int,
string: String,
long: Long
)
Config for this example:
align.openParenDefnSite = false
align.openBracketDefnSite
Since v3.0.4.
If set explicitly, will be used for the left bracket in type parameters,
instead of align.openParenDefnSite.
align.openParenTupleSite
This parameter controls aligning of tuples to their opening parenthesis. If not
specified, will use the value of align.openParenCallSite.
Since v3.0.0.
object a {
foo(
bar,
baz
)
(bar,
baz
)
}
object a {
foo(bar, baz)
(bar, baz)
}
Config for this example:
maxColumn = 10
align.openParenCallSite = false
align.openParenTupleSite = true
object a {
foo(bar,
baz
)
(bar,
baz
)
}
object a {
foo(bar, baz)
(bar, baz)
}
Config for this example:
maxColumn = 10
align.openParenCallSite = true
align.beforeOpenParenXxxSite
Aligns parameter groups (not parameters within a group) if using
newlines.beforeOpenParenXxxSite.
Requires align.closeParenSite.
Since v3.3.2.
# Defaults
align.beforeOpenParenCallSite = false
align.beforeOpenParenDefnSite = false
align.stripMargin
See assumeStandardLibraryStripMargin.
Default: align.stripMargin = true
This functionality is enabled in all presets except align.preset=none where it
was disabled since the parameter's introduction in v2.5.0.
align.multiline
If this flag is set, when alignment is applied, multiline statements will not be excluded from search of tokens to align.
Since v2.5.0.
Default: align.multiline = false
for {
a <- aaa
bbb <- bb
cccccc <- c {
3
}
dd <- ddddd
} yield ()
for {
a <- aaa
bbb <- bb
cccccc <- c {
3
}
dd <- ddddd
} yield ()
Config for this example:
align.preset = more
align.multiline = true
for {
a <- aaa
bbb <- bb
cccccc <- c {
3
}
dd <- ddddd
} yield ()
for {
a <- aaa
bbb <- bb
cccccc <- c {
3
}
dd <- ddddd
} yield ()
Config for this example:
align.preset = more
align.multiline = false
align.allowOverflow
If this flag is set, as long as unaligned lines did not overflow, we will not check
whether alignment causes any lines to overflow the maxColumn setting.
Since v3.0.0.
Default: align.allowOverflow = false
It's also enabled by default in align.preset = most.
align.inInterpolation
If this flag is set, and breaks within interpolated code are allowed
(see newlines.inInterpolation, then the
interpolated code and the closing } will be indented relative to the opening ${.
Since v3.4.0.
Default: align.inInterpolation = false
Keep in mind that this option might lead to line overflow via "stacking":
object a {
s"""
|foo1 ${
quxQux(
bazBaz,
barBar
)
} foo2 ${
quxQux(
bazBaz,
barBar
)
} foo3 ${
quxQux(
bazBaz,
barBar
)
} foo4
|""".stripMargin
}
object a {
s"""
|foo1 ${quxQux(bazBaz, barBar)} foo2 ${quxQux(bazBaz, barBar)} foo3 ${quxQux(bazBaz, barBar)} foo4
|""".stripMargin
}
Config for this example:
maxColumn = 30
align.inInterpolation = true
newlines.inInterpolation = oneline
vs
object a {
s"""
|foo1 ${
quxQux(bazBaz, barBar)
} foo2 ${
quxQux(bazBaz, barBar)
} foo3 ${
quxQux(bazBaz, barBar)
} foo4
|""".stripMargin
}
object a {
s"""
|foo1 ${quxQux(bazBaz, barBar)} foo2 ${quxQux(bazBaz, barBar)} foo3 ${quxQux(bazBaz, barBar)} foo4
|""".stripMargin
}
Config for this example:
maxColumn = 30
align.inInterpolation = false
newlines.inInterpolation = oneline
align.delayUntilSpace
If this flag is set, the formatter will not forcefully pull apart two successive non-whitespace tokens that would otherwise be formatted without a space between them.
Instead, the extra alignment spaces will be added to the next space on the same line.
Since v3.7.13. Prior to that, this behaviour was always enabled.
Default: align.delayUntilSpace = true
object a {
def meeethod1(pram1: AnyRef): Any =
???
def methd2(paaaaaram2: Any): Any = ???
def meth3(param333333: Any): Any = ???
def md4(param4: Any): Any = ???
}
object a {
def meeethod1(pram1: AnyRef): Any = ???
def methd2(paaaaaram2: Any): Any = ???
def meth3(param333333: Any): Any = ???
def md4(param4: Any): Any = ???
}
Config for this example:
align.preset = more
align.delayUntilSpace = true
align.tokens."+" = [ { code = ":" }, { code = "(" }, { code = ")" }, { code = "=" } ]
vs
object a {
def meeethod1(pram1: AnyRef): Any =
???
def methd2(paaaaaram2: Any): Any = ???
def meth3(param333333: Any): Any = ???
def md4 (param4 : Any): Any = ???
}
object a {
def meeethod1(pram1: AnyRef): Any = ???
def methd2(paaaaaram2: Any): Any = ???
def meth3(param333333: Any): Any = ???
def md4(param4: Any): Any = ???
}
Config for this example:
align.preset = more
align.delayUntilSpace = false
align.tokens."+" = [ { code = ":" }, { code = "(" }, { code = ")" }, { code = "=" } ]
Newlines
The newlines.* options are used to configure when and where scalafmt should
insert newlines.
You might be interested in the Vertical Multiline section.
newlines.source
Since v2.5.0.
This parameter controls the general approach to line breaks, and whether to take into account existing newlines in the source. The default value (if the parameter is not specified) is the classic, original way. Below are the alternatives.
newlines.source=keep
This approach attempts to preserve line breaks in the input whenever possible.
Keep in mind that other parameters will take precedence if they require (or prohibit) breaks in specific places (and that includes their default values).
For instance, default for
binPack.callSitewill not allow multiple arguments per line in a multiline expression.
newlines.source=fold,unfold
These two approaches completely ignore existing line breaks, except around comments and blank lines (i.e., multiple consecutive newlines).
Might require increasing optimizer limits, to avoid SearchStateExploded exceptions.
fold attempts to remove line breaks whenever possible resulting in a more
horizontal, or vertically compact look.
unfold, on the other hand, is intended for those who prefer a more vertical,
or horizontally compact look.
Both settings attempt to play nice with other parameters, but some combinations are prohibited and will result in an error.
newlines.topLevelStatementBlankLines
Since v3.0.0.
This parameter controls when to add blank lines before and/or after a top-level statement (a member of a package or template; nesting is allowed but not within a block). Special cases:
- the rules do not directly apply to package statements at the top of the source file; however, if this parameter is non-empty, there will be at least one blank line before the first non-package statement, and possibly more if the rules match that statement
- end markers are handled through a setting for the statement they mark
- imports and exports are processed as a group of consecutive statements
- also see Newlines around package or template body
This parameter might reduce the number of blank lines but will not eliminate them completely unless corresponding value is negative.
Each entry on this list consists of the following fields (except blanks, all are used to match
whether the rule should apply):
regex- a regular expression to match the type of the statement
- if unspecified, will match all valid statements
- see align.tokens for instructions on how to find the type
maxNestand (since v3.1.2)minNest- basically, limits indentation level (not actual indentation) of a statement
- unindented statements (under source-level unindented package) have nest level of 0, those under them are 1 etc.
- if unspecified, will match any nesting level
minBreaks(default: 1)- sets the minimum number of line breaks between the first and last line of a statement (i.e., one less than the number of lines the statement spans).
- for instance,
minBreaks=0will apply to all statements, whereas 1 will require at least one line break (that is, a multi-line statement).
- (since v3.8.3)
minBlankGapsandmaxBlankGaps- sets the range on the number of blank-line gaps (regardless of the number of lines in each gap) that would have been output prior to this rule.
- if a limit is unspecified, will not constrain on the corresponding end
blanks- if omitted while the entry matches, serves to exclude another entry
before: number of lines to be added before a matching statement; if negative (v3.0.1), lines will be removed unless before a commentafter: number of lines to be added after a matching statement- for instance, if a
packagematches, this controls how many lines need to be added after all statements of a package, not after the first line which declares the package name - same logic as above applies to negative values
- for instance, if a
beforeEndMarker:- end markers themselves will not be matched against any rule; blanks before
them will come from
beforeEndMarkerand blanks after fromafter
- end markers themselves will not be matched against any rule; blanks before
them will come from
beforeAllandafterAll(v3.0.1): if set explicitly, replaces the respectivebeforeoraftervalue before the first or after the last statement of a template or indented package; otherwise, thebeforeoraftervalue will be capped at 1- can be specified as a single integer, to set just
beforeandafterto the same value:
// these two are equivalent
newlines.topLevelStatementBlankLines = [
{ blanks { before = 1, after = 1, beforeEndMarker = 0 } }
]
newlines.topLevelStatementBlankLines = [
{ blanks = 1 }
]
If you'd like to override or exclude some cases, add them explicitly:
newlines.topLevelStatementBlankLines = [
{ maxNest = 0, blanks = 2 } // uses 2 blanks for all unindented statements
{ regex = "^Import" } // excludes import groups; blanks are not specified
]
If multiple entries match a statement, an entry with the lowest
minBreakswill be selected. Since we'll be adding lines, this will increase the span of the statement and might potentially lead to another entry, with a higherminBreaks, to match as well, which is undesirable.
Default: newlines.topLevelStatementBlankLines = []
import org.scalafmt
package core { // no newline added here
class C1 {}
object O { // nor here
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
} // nor here
class C2 {}
}
import org.scalafmt
package core { // no newline added here
class C1 {}
object O { // nor here
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
} // nor here
class C2 {}
}
Config for this example:
newlines.topLevelStatementBlankLines = []
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
Config for this example:
newlines.topLevelStatementBlankLines = [
{
blanks { before = 1 }
}
]
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
Config for this example:
newlines.topLevelStatementBlankLines = [
{
blanks { after = 1 }
}
]
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {}
object O {
val x1 = 1
val x2 = 2
def A = "A"
def B = "B"
}
class C2 {}
}
Config for this example:
newlines.topLevelStatementBlankLines = [
{
blanks = 1
}
]
package core {
object O {
val x1 = 1
val x2 = 2
def A =
"A"
def B = {
"B"
}
}
}
package core {
object O {
val x1 = 1
val x2 = 2
def A =
"A"
def B = {
"B"
}
}
}
Config for this example:
newlines.topLevelStatementBlankLines = [
{
minBreaks = 0
blanks = 1
}
]
import org.scalafmt
package core {
object O {
val x1 = 1
val x2 = 2
def A =
"A"
def B = {
"B"
}
}
}
import org.scalafmt
package core {
object O {
val x1 = 1
val x2 = 2
def A =
"A"
def B = {
"B"
}
}
}
Config for this example:
newlines.topLevelStatementBlankLines = [
{
minBreaks = 2
blanks = 1
}
]
Newlines around package or template body
Since v3.0.0.
This group of parameters controls whether to enforce a blank line before the first or after the last statement of a package or template (i.e., body of a class, object, trait, enum).
These parameters will not cause any blank lines to be removed.
newlines.topLevelBodyIfMinStatements
topLevelBodyIfMinStatements can be before and/or after, while
topLevelBodyMinStatements limits when the rule is applied.
# Defaults
newlines.topLevelBodyIfMinStatements = []
newlines.topLevelBodyMinStatements = 2
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
Config for this example:
newlines.topLevelBodyIfMinStatements = []
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
Config for this example:
newlines.topLevelBodyIfMinStatements = [before]
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
Config for this example:
newlines.topLevelBodyIfMinStatements = [after]
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
import org.scalafmt
package core {
class C1 {
def one = 1
}
object O1 {
val one = 1
def two = 2
}
class C2 {}
}
Config for this example:
newlines.topLevelBodyIfMinStatements = [before,after]
newlines.beforeTemplateBodyIfBreakInParentCtors
This parameter will force a blank line before the first statement of a template body if the
token before extends and the { (or, in scala3, :) token are not on the same line.
Default: newlines.beforeTemplateBodyIfBreakInParentCtors = false
package core {
class C1 extends S { // no breaks between "C1" and "{"
def one = 1
}
class C1(
param: Int
) extends S { // no breaks between ")" and "{"
def one = 1
}
class C1 extends S { // no breaks between "C1" and "=>"
self =>
def one = 1
}
class C1
extends S { // break between "C1" and "{"
def one = 1
}
}
package core {
class C1 extends S { // no breaks between "C1" and "{"
def one = 1
}
class C1(
param: Int
) extends S { // no breaks between ")" and "{"
def one = 1
}
class C1 extends S { // no breaks between "C1" and "=>"
self =>
def one = 1
}
class C1
extends S { // break between "C1" and "{"
def one = 1
}
}
Config for this example:
newlines.source = keep
newlines.beforeTemplateBodyIfBreakInParentCtors = true
package core {
class C1 extends S { // no breaks between "C1" and "{"
def one = 1
}
class C1(
param: Int
) extends S { // no breaks between ")" and "{"
def one = 1
}
class C1 extends S { // no breaks between "C1" and "=>"
self =>
def one = 1
}
class C1
extends S { // break between "C1" and "{"
def one = 1
}
}
package core {
class C1 extends S { // no breaks between "C1" and "{"
def one = 1
}
class C1(
param: Int
) extends S { // no breaks between ")" and "{"
def one = 1
}
class C1 extends S { // no breaks between "C1" and "=>"
self =>
def one = 1
}
class C1
extends S { // break between "C1" and "{"
def one = 1
}
}
Config for this example:
newlines.source = keep
newlines.beforeTemplateBodyIfBreakInParentCtors = false
newlines.beforeMultiline
Since v2.7.0
This parameter controls whether to force a new line before a multi-line body of
case/if/def/val and how to format it if the space is allowed. (For
additional control with assignment expressions, please also see
newlines.forceBeforeMultilineAssign below.)
It accepts the same values as newlines.source (and defaults
to that parameter's setting).
NB: for breaks before parameters of a multi-line lambda, use multiline with
newlines.beforeCurlyLambdaParams.
a match {
// had space after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
Config for this example:
newlines.beforeMultiline = unfold
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
}
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
Config for this example:
newlines.beforeMultiline = fold
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
Config for this example:
newlines.beforeMultiline = keep
a match {
// had space after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
a match {
// had space after "=>"
case a => if (step != 0)
d.name should be("dir" + step)
// had newline after "=>"
case a =>
if (step != 0)
d.name should be("dir" + step)
}
Config for this example:
# newlines.beforeMultiline = classic
newlines.forceBeforeMultilineAssign
Since v3.0.0
This section controls whether to force a break before a multi-line body of an
assignment expression unless it can be formatted on a single line (or is enclosed
in braces). By default, the rule is disabled. It takes precedence
over newlines.beforeMultiline settings.
It can take the following values:
never: the rule is disabledany: applies to any assignment expression (def, assignment to avar, default value of a method parameter, etc.)def: applies only to definitions which can potentially be parameterized (def,macro,givenalias, etc.)anyMember: applies to members of aclass/trait/objecttopMember: applies to members of aclass/trait/objectwhich itself can only be nested within a sequence ofclass/trait/objectdefinitions
It replaces deprecated newlines parameters beforeMultilineDef=unfold and
alwaysBeforeMultilineDef=true which, if this parameter is not set,
map to def.
class A {
// break, allows params (even if it doesn't define any)
def foo =
func(
foo,
bar
)
// no break, doesn't allow params
val foo = func(
foo,
bar
)
def foo = {
def a =
func(
foo,
bar
)
val a = func(
foo,
bar
)
}
}
class A {
// break, allows params (even if it doesn't define any)
def foo = func(foo, bar)
// no break, doesn't allow params
val foo = func(foo, bar)
def foo = {
def a = func(foo, bar)
val a = func(foo, bar)
}
}
Config for this example:
maxColumn = 17
newlines.forceBeforeMultilineAssign = def
class A {
class B {
// break, a top member
def foo = func(
foo,
bar
)
// break, a top member
val foo = func(
foo,
bar
)
}
def foo = {
// no break, not a member
def a = func(
foo,
bar
)
// no break, not a member
val a = func(
foo,
bar
)
new A with B {
// no break, not a top member
def foo =
func(
foo,
bar
)
// no break, not a top member
val foo =
func(
foo,
bar
)
}
}
}
class A {
class B {
// break, a top member
def foo = func(foo, bar)
// break, a top member
val foo = func(foo, bar)
}
def foo = {
// no break, not a member
def a = func(foo, bar)
// no break, not a member
val a = func(foo, bar)
new A with B {
// no break, not a top member
def foo = func(foo, bar)
// no break, not a top member
val foo = func(foo, bar)
}
}
}
Config for this example:
maxColumn = 19
newlines.forceBeforeMultilineAssign = topMember
// all disabled, no breaks
class A {
def foo = func(
foo,
bar
)
val foo = func(
foo,
bar
)
def foo = {
def a = func(
foo,
bar
)
val a = func(
foo,
bar
)
}
}
// all disabled, no breaks
class A {
def foo = func(foo, bar)
val foo = func(foo, bar)
def foo = {
def a = func(foo, bar)
val a = func(foo, bar)
}
}
Config for this example:
maxColumn = 17
newlines.forceBeforeMultilineAssign = never
newlines.forceBeforeAssign
Since v3.5.9
Default: newlines.forceBeforeAssign = "never"
This parameter takes precedence over
newlines.forceBeforeMultilineAssign
and uses the same values. The difference is, the rule forces a newline before a
matching assignment expression whether or not it can be formatted on a single line.
newlines.beforeTypeBounds
Since v3.0.0
This parameter controls formatting of bounds of type parameters: upper <:,
lower >:, view <%, and context : bounds. It accepts the
same values as newlines.source.
classic: simply allows no line breaks (default; can't be specified)keep: preserves a no-break if the next bound fits on the linefold: uses a no-break if the next bound fits on the lineunfold: puts all bounds on the same line, or breaks before each
newlines.alwaysBeforeElseAfterCurlyIf
Default: newlines.alwaysBeforeElseAfterCurlyIf = false
if (someCond) {
foo()
}
else {
bar()
}
if (someCond) {
foo()
} else {
bar()
}
Config for this example:
newlines.alwaysBeforeElseAfterCurlyIf = true
if (someCond) {
foo()
} else {
bar()
}
if (someCond) {
foo()
}
else {
bar()
}
Config for this example:
newlines.alwaysBeforeElseAfterCurlyIf = false
newlines.beforeCurlyLambdaParams
This parameter controls whether a newline is forced between the opening curly
brace and the parameters of a lambda or partial function. Added in 2.7.0,
replacing boolean alwaysBeforeCurlyBraceLambdaParams (removed in 3.4.0).
Default: newlines.beforeCurlyLambdaParams = "never"
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break on arrow since case doesn't fit on a line
x.zipWithIndex.map { case (c, i) =>
s"$c -> $i"
}
x.zipWithIndex.map { case (c, i) =>
s"$c -> $i (long comment)"
}
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break on arrow since case doesn't fit on a line
x.zipWithIndex.map { case (c, i) => s"$c -> $i" }
x.zipWithIndex.map { case (c, i) => s"$c -> $i (long comment)" }
Config for this example:
newlines.beforeCurlyLambdaParams = never
// should break on brace, though fits on the same line
x.map {
x => s"${x._1} -> ${x._2}"
}
x.map {
case (c, i) => s"$c -> $i"
}
x.zipWithIndex.map {
case (c, i) =>
s"$c -> $i (long comment)"
}
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map {
x =>
s"${x._1} -> ${x._2} (long comment)"
}
// should break on brace, though fits on the same line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
x.zipWithIndex.map { case (c, i) => s"$c -> $i (long comment)" }
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map { x => s"${x._1} -> ${x._2} (long comment)" }
Config for this example:
newlines.beforeCurlyLambdaParams = always
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break on brace as lambda doesn't fit on the same line
x.zipWithIndex.map {
x => s"${x._1} -> ${x._2}"
}
x.zipWithIndex.map {
case (c, i) => s"$c -> $i"
}
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map {
x =>
s"${x._1} -> ${x._2} (long comment)"
}
x.zipWithIndex.map {
case (c, i) =>
s"$c -> $i (long comment)"
}
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break on brace as lambda doesn't fit on the same line
x.zipWithIndex.map { x => s"${x._1} -> ${x._2}" }
x.zipWithIndex.map { case (c, i) => s"$c -> $i" }
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map { x => s"${x._1} -> ${x._2} (long comment)" }
x.zipWithIndex.map { case (c, i) => s"$c -> $i (long comment)" }
Config for this example:
newlines.beforeCurlyLambdaParams = multiline
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break after arrow as lambda doesn't fit on the same line
x.zipWithIndex.map { x =>
s"${x._1} -> ${x._2}"
}
x.zipWithIndex.map { x =>
s"${x._1} -> ${x._2} (long comment)"
}
// should break on brace as lambda doesn't fit on the same line
x.zipWithIndex.map {
case (c, i) => s"$c -> $i"
}
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map {
case (c, i) =>
s"$c -> $i (long comment)"
}
// should keep one-line
x.map { x => s"${x._1} -> ${x._2}" }
x.map { case (c, i) => s"$c -> $i" }
// should break after arrow as lambda doesn't fit on the same line
x.zipWithIndex.map { x => s"${x._1} -> ${x._2}" }
x.zipWithIndex.map { x => s"${x._1} -> ${x._2} (long comment)" }
// should break on brace as lambda doesn't fit on the same line
x.zipWithIndex.map { case (c, i) => s"$c -> $i" }
// should break on brace and arrow as lambda doesn't fit on a line
x.zipWithIndex.map { case (c, i) => s"$c -> $i (long comment)" }
Config for this example:
newlines.beforeCurlyLambdaParams = multilineWithCaseOnly
newlines.afterCurlyLambdaParams
This parameter controls handling of newlines after the arrow following the parameters of a curly brace lambda or partial function, and whether a space can be used for one-line formatting of the entire function body (if allowed but the body doesn't fit, a break is always forced).
This parameter was renamed in 2.7.0 from afterCurlyLambda, for clarity and
consistency with beforeCurlyLambdaParams defined above.
Default: newlines.afterCurlyLambdaParams = "never"
// remove all blank lines if any
// one-line formatting is allowed
something.map { x => f(x) }
something.map { x => f(x) }
// remove all blank lines if any
// one-line formatting is allowed
something.map { x =>
f(x)
}
something.map { x => f(x) }
Config for this example:
newlines.afterCurlyLambdaParams = squash
// remove all blank lines if any
// one-line formatting depends on newlines.source:
// yes for fold; no for unfold; otherwise, only if there was no break
something.map { x =>
f(x)
}
something.map { x => f(x) }
// remove all blank lines if any
// one-line formatting depends on newlines.source:
// yes for fold; no for unfold; otherwise, only if there was no break
something.map { x =>
f(x)
}
something.map { x => f(x) }
Config for this example:
newlines.afterCurlyLambdaParams = never
// if blank lines are present, keep only one
// one-line formatting depends on newlines.source:
// if no blank for fold; no for unfold; otherwise, only if there was no break
something.map { x =>
f(x)
}
something.map { x => f(x) }
// if blank lines are present, keep only one
// one-line formatting depends on newlines.source:
// if no blank for fold; no for unfold; otherwise, only if there was no break
something.map { x =>
f(x)
}
something.map { x => f(x) }
Config for this example:
newlines.afterCurlyLambdaParams = keep
// ensure a single blank line
// one-line formatting is not allowed
something.map { x =>
f(x)
}
something.map { x =>
f(x)
}
// ensure a single blank line
// one-line formatting is not allowed
something.map { x =>
f(x)
}
something.map { x => f(x) }
Config for this example:
newlines.afterCurlyLambdaParams = always
newlines.implicitParamListModifierXXX
These parameters control newlines around implicit parameter list modifier.
Since v2.5.0.
Newlines around using parameter and argument list modifier
using soft keyword was introduced in Scala 3 and is supported in scalafmt.
Besides parameter lists, using can also be used with argument lists hence
provided rules will then also be applied to them.
Since v3.0.0.
The parameters defined below can also be accessed via their alternative names where
implicit is replaced with using (for instance, newlines.usingParamListModifierPrefer).
Whichever naming you use, formatting will be applied to both implicit and using.
Prefer After (default)
Prefers newline after
implicit. Newline will be added unless the entire implicit parameter list fits on a line, or config style is false. Newline can also be added before if the keyword itself would overflow the line.
def format(code: String, age: Int)(implicit
ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
newlines.implicitParamListModifierPrefer = after
Prefer Before
Prefers newline before
implicit. Newline will not be added if the entire implicit parameter list fits on a line.
def format(code: String, age: Int)(
implicit ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
newlines.implicitParamListModifierPrefer = before
Force Before
If set, forces newline before
implicit(even if the parameter list would fit on one line). Otherwise, newline can still be added if the keyword would overflow the line.
def format(code: String, age: Int)(
implicit ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
newlines.implicitParamListModifierForce = [before]
Force After
If set, forces newline after
implicit(even if the parameter list would fit on one line). Otherwise, newline can still be added unlessbeforeis true, or the entire implicit parameter list fits on a line, or config style is false.
def format(code: String, age: Int)(implicit
ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
newlines.implicitParamListModifierForce = [after]
Force both before and after
def format(code: String, age: Int)(
implicit
ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
newlines.implicitParamListModifierForce = [before,after]
Implicit with newlines.configStyle.defnSite.prefer
While config-style normally requires a newline after the opening parenthesis,
postponing that break until after the implicit keyword is allowed if other
parameters require keeping this keyword attached to the opening brace.
Therefore, any of the parameters described in this section will take precedence
even when newlines.configStyle.defnSite.prefer = true is used.
newlines.afterInfix
Since v2.5.0.
This parameter (and its companions) controls formatting around infix expressions.
The default value depends on
newlines.source(see below).
newlines.afterInfix=keep
This approach preserves line breaks in the input. This is the original
behaviour, and default for newlines.source=classic,keep.
newlines.afterInfix=many,some
These approaches completely ignore existing newlines around infix, always use
a space before an infix operator and occasionally break after it. some is
default for newlines.source=fold, and many for newlines.source=unfold.
Might require increasing optimizer limits, to avoid SearchStateExploded exceptions.
some will introduce fewer line breaks than many. Both will attempt to break
after
higher-precedence operators,
and both will always break before an expression enclosed in matching
parentheses.
newlines.afterInfixMaxCountPerFile
Default: newlines.afterInfixMaxCountPerFile = 500
If the total number of infix operations in the entire file exceeds
newlines.afterInfixMaxCountPerFile, the formatter automatically switches to
newlines.afterInfix=keep for this file.
newlines.afterInfixMaxCountPerExprForSome
Default: newlines.afterInfixMaxCountPerExprForSome = 10
If newlines.afterInfix is set to some and the number of infix operations in
a given expression sequence (top-level or enclosed in parens/braces) exceeds
newlines.afterInfixMaxCountPerExprForSome, the formatter switches to many
for that sequence only.
newlines.afterInfixBreakOnNested
Default: newlines.afterInfixBreakOnNested = false
If enabled, will force line breaks around a nested parenthesized sub-expression in a multi-line infix expression.
newlines.avoidForSimpleOverflow
A list parameter (of comma-separated flags), with possible flags described
below. These flags relax formatting rules to allow occasional line overflow
(i.e., when line exceeds maxColumn) in simple cases instead of introducing a
newline.
Default: newlines.avoidForSimpleOverflow = []
newlines.avoidForSimpleOverflow=[tooLong]
Since v2.6.0.
This flag tries to avoid introducing a newline if the line would overflow even with a newline.
object Example {
foo_bar_baz("the quick brown fox jumps over the lazy dog") {
println("")
}
foo_bar_baz(
"the quick brown fox jumps over a dog") {
println("")
}
}
object Example {
foo_bar_baz("the quick brown fox jumps over the lazy dog") {
println("")
}
foo_bar_baz("the quick brown fox jumps over a dog") {
println("")
}
}
Config for this example:
maxColumn = 50
danglingParentheses.callSite = false
newlines.avoidForSimpleOverflow = [tooLong]
newlines.avoidForSimpleOverflow=[punct]
Since v2.6.0.
This flag tries to avoid a newline if the line would overflow only because of trailing punctuation (non-alphanum symbols of length 1).
With the flag set:
class Engine[TD, EI, PD, Q, P, A](
val dataSourceClassMap: Map[String, Class[_ <: BaseDataSource[TD, EI, Q, A]]]
) {}
class Engine[TD, EI, PD, Q, P, A](
val dataSourceClassMap: Map[
String,
Class[_ <: BaseDataSource[TD, EI, Q, A]]]) {}
Config for this example:
maxColumn = 80
newlines.avoidForSimpleOverflow = [punct]
newlines.avoidForSimpleOverflow=[slc]
Since v3.0.0.
This flag tries to avoid a newline if the line would overflow only because of
trailing single-line comment (one which starts with //).
Also, since v3.3.1, this includes a trailing /* ... */ without embedded breaks.
intercept[TestException] {
val ct = Thread.currentThread() // comment
}
intercept[TestException] {
val ct = Thread.currentThread() // comment
}
Config for this example:
maxColumn = 40
newlines.avoidForSimpleOverflow = [slc]
intercept[TestException] {
val ct =
Thread.currentThread() // comment
}
intercept[TestException] {
val ct = Thread.currentThread() // comment
}
Config for this example:
maxColumn = 40
newlines.avoidForSimpleOverflow = []
newlines.avoidInResultType
If true, newlines in definition result type will only be used if formatting
without them is impossible. This parameter was added in 2.7.0, replacing
neverInResultType.
Default: newlines.avoidInResultType = false
// no newlines in result type
def permissionState(
a: A = js.native
): js.Promise[PushPermissionState] = js.native
// no newlines in result type
val permissionState
: js.Promise[PushPermissionState] = js.native
// can't format without newlines
implicit protected val td: TildeArrow {
type Out = RouteTestResult
} = TildeArrow.injectIntoRoute
// no newlines in result type
def permissionState(a: A = js.native): js.Promise[PushPermissionState] = js.native
// no newlines in result type
val permissionState: js.Promise[PushPermissionState] = js.native
// can't format without newlines
implicit protected val td: TildeArrow {
type Out = RouteTestResult } = TildeArrow.injectIntoRoute
Config for this example:
maxColumn = 40
newlines.avoidInResultType = true
newlines.neverBeforeJsNative = true
newlines.sometimesBeforeColonInMethodReturnType
If true, in some rare cases a newline could be added before the colon which delimits the return type in a method signature.
Normally, this will only happen if formatting otherwise will lead to line overflow or the return type spanning too many lines.
Also, see newlines.beforeOpenParenXxxSite below.
Default: newlines.sometimesBeforeColonInMethodReturnType = true
def someVeryLongMethodName
: Map[String, String] = ???
def someVeryLongMethodName: Map[String, String] = ???
Config for this example:
maxColumn = 40
newlines.sometimesBeforeColonInMethodReturnType = true
newlines.beforeOpenParenXxxSite
This parameter, if enabled, causes a parameter group which doesn't fit on the same line to be formatted separately, with a newline just before the opening parenthesis.
Since v3.0.0.
Use null (default) to disable it. Otherwise, takes the same values as
newlines.source (or explicit source to mirror the current
value of newlines.source).
Specific formatting depends on the value: for keep, will preserve a newline;
for unfold, will format every parameter group separately unless all fit on a
single line; fold will use a more compact formatting.
beforeOpenParenDefnSiteapplies to definitions (methods, ctors, macros, etc.)beforeOpenParenCallSiteapplies to invocations (method calls) and is only supported for scala3
Additional nuances:
- if
newlines.sometimesBeforeColonInMethodReturnTypeis true, a newline will be added before the colon unless the entire signature fits on a line (except when set tokeep). - if the corresponding
align.openParenXxxSiteis true, multi-line parameters will start on the same line as the opening parenthesis and align; otherwise, formatting will use a newline and an appropriate continuation indent. - if the corresponding
align.beforeOpenParenXxxSiteis true, when the first parameter group starts without a line break, subsequent parameter groups will be aligned to it.
# Defaults
newlines.beforeOpenParenDefnSite = null
newlines.beforeOpenParenCallSite = null
def fooFunc
(foo1: String)
(foo2: String,
foo3: String
)
: String = ???
val res = fooFunc
("foo1")
("foo2", "foo3")
def fooFunc(foo1: String)(foo2: String, foo3: String): String = ???
val res = fooFunc("foo1")("foo2", "foo3")
Config for this example:
maxColumn = 20
runner.dialect = scala3 // for CallSite
align.openParenDefnSite = true
newlines {
beforeOpenParenDefnSite = fold
beforeOpenParenCallSite = unfold
sometimesBeforeColonInMethodReturnType = true
}
def fooFunc
(foo1: String)
(
foo2: String,
foo3: String
): String = ???
val res = fooFunc
("foo1")
("foo2", "foo3")
def fooFunc(foo1: String)(foo2: String, foo3: String): String = ???
val res = fooFunc("foo1")("foo2", "foo3")
Config for this example:
maxColumn = 20
runner.dialect = scala3 // for CallSite
align.openParenCallSite = true
newlines {
beforeOpenParenDefnSite = unfold
beforeOpenParenCallSite = fold
sometimesBeforeColonInMethodReturnType = false
}
newlines.inInterpolation
This parameter controls how to format spliced scala code within string constants
(e.g., s"...", etc). Also see align.inInterpolation.
Default: newlines.inInterpolation = "allow"
Since v3.4.0.
allow: allows breaks within spliced code (original)- this option will not prevent line overflow even if
${is within bounds, because this option doesn't allow breaking right after${
- this option will not prevent line overflow even if
avoid: attemps to avoid breaks within the spliced code, regardless of line overflowoneline: formats the splice on a single line, or breaks after${if overflows
newlines.ignoreInSyntax
The formatter frequently chooses between adding a newline and continuing the same line but either prohibiting or heavily discouraging subsequent newlines between tokens, to fit the rest of the expression on the same line.
However, in many cases and, for historical reasons, intentionally, newlines within tokens have been frequently ignored, leading to "single-line" blocks which actually span multiple lines.
This boolean parameter now allows controlling whether to ignore newlines found in syntax of strings or other possibly multi-line tokens when newlines are otherwise prohibited or undesirable (such as for single-line formatting).
Default: newlines.ignoreInSyntax = true
Since v3.7.13. Prior to that, this behaviour was always enabled.
// ignores newline in string, pretends everything fits on one line
println(s"""${1}
""".stripMargin)
// ignores newline in string, pretends everything fits on one line
println(s"""${1}
""".stripMargin
)
Config for this example:
newlines.ignoreInSyntax = true
// detects newline in string, forces proper multi-line formatting
println(
s"""${1}
""".stripMargin
)
// detects newline in string, forces proper multi-line formatting
println(s"""${1}
""".stripMargin
)
Config for this example:
newlines.ignoreInSyntax = false
newlines.annotation
This boolean parameter controls newlines after annotations. Prior to v3.8.4,
was called optIn.annotationNewlines.
Default: newlines.annotation = true
Its behaviour depends on newlines.source:
newlines.annotation = true:newlines.source=fold: allows space before another annotationnewlines.source=unfold: forces break- otherwise: preserves space before or after an annotation
newlines.annotation = false:newlines.source=fold: allows space before a keyword or another annotationnewlines.source=unfold: allows space before another annotation- otherwise: allows space before a keyword
newlines.selfAnnotation
If enabled, this parameter forces a break after a self annotation in a template if
newlines.source=fold/unfold, and preserves it otherwise.
Prior to v3.8.4, was called optIn.selfAnnotationNewline.
Default: newlines.selfAnnotation = true
Newlines: danglingParentheses
When enabled, these parameters will force a break before the closing parenthesis
when there's a break after the opening parenthesis (unless otherwise specified).
While this parameter is not technically under the newlines section, it
logically belongs there.
danglingParentheses.defnSite
Default: danglingParentheses.defnSite = true
object a {
// defnSite
def method(
a: Int,
b: String
): Boolean
// callSite
method(
argument1,
argument2)
}
object a {
// defnSite
def method(a: Int, b: String): Boolean
// callSite
method(argument1, argument2)
}
Config for this example:
danglingParentheses.defnSite = true
danglingParentheses.callSite = false
maxColumn=25
danglingParentheses.callSite
Default: danglingParentheses.callSite = true
object a {
// defnSite
def method(
a: Int,
b: String): Boolean
// callSite
method(
argument1,
argument2
)
}
object a {
// defnSite
def method(a: Int, b: String): Boolean
// callSite
method(argument1, argument2)
}
Config for this example:
danglingParentheses.defnSite = false
danglingParentheses.callSite = true
maxColumn=25
danglingParentheses.bracketXxxSite
These control the bracketed type clauses in definitions and method calls and by default are equal to their non-bracket counterparts.
object a {
// defnSite
def method[A <: Any, B]
: Boolean
// callSite
method[
String,
SomeOtherType]
}
object a {
// defnSite
def method[A <: Any, B]: Boolean
// callSite
method[String, SomeOtherType]
}
Config for this example:
danglingParentheses.defnSite = true
danglingParentheses.callSite = true
danglingParentheses.bracketDefnSite = false
danglingParentheses.bracketCallSite = false
maxColumn=25
object a {
// defnSite
def method[A <: Any, B]
: Boolean
// callSite
method[
String,
SomeOtherType
]
}
object a {
// defnSite
def method[A <: Any, B]: Boolean
// callSite
method[String, SomeOtherType]
}
Config for this example:
danglingParentheses.defnSite = false
danglingParentheses.callSite = false
danglingParentheses.bracketDefnSite = true
danglingParentheses.bracketCallSite = true
maxColumn=25
danglingParentheses.ctrlSite
Since v2.5.0.
Forces dangling on open/close parens around control structures (if, while,
for) when line breaks must occur.
For optional braces in scala3, this parameter also controls whether to break
before then or do in a multi-line condition.
Default: danglingParentheses.ctrlSite = true
if (something) {
// nothing
}
if (
something_else
) {
// nothing
}
if (something) {
// nothing
}
if (something_else) {
// nothing
}
Config for this example:
danglingParentheses.ctrlSite = true
maxColumn=20
danglingParentheses.tupleSite
This parameter controls dangling of closing parentheses in tuples. If not
specified, will use the value of danglingParentheses.callSite.
Since v3.0.0.
val i =
( // hello scalafmt, please don't blow up
(1, 2),
1,
3,
4,
4
)
val i =
( //hello scalafmt, please don't blow up
(1, 2),
1,
3,
4,
4)
Config for this example:
danglingParentheses.tupleSite = true
val i =
( // hello scalafmt, please don't blow up
(1, 2),
1,
3,
4,
4)
val i =
( //hello scalafmt, please don't blow up
(1, 2),
1,
3,
4,
4)
Config for this example:
danglingParentheses.tupleSite = false
danglingParentheses.exclude
Since v2.5.0.
When the appropriate danglingParentheses flag (e.g., defnSite) has been set,
this parameter can be used to limit contexts where dangling is applied
(currently, class, trait, enum, extension and def are supported).
For backwards compatibility, the default depends on whether
Vertical Multiline mode is used. If it is, the default is
[class, trait]; otherwise, it's empty.
def other(a: String, b: String)(
c: String,
d: String) = a + b + c
other(a, b)(c, d)
def other(a: String, b: String)(c: String, d: String) = a + b + c
other(a, b)(c, d)
Config for this example:
indent.defnSite = 2
danglingParentheses.defnSite = true
danglingParentheses.exclude = [def]
Newlines: Config-style formatting
This formatting applies to argument clauses in method calls or parameter
clauses in definitions. It outputs a newline after the opening parenthesis
(or after the implicit keyword) and a newline before the closing parenthesis,
with arguments or parameters listed one per line.
newlines.configStyle.xxxSite
The xxxSite portion refers to:
fallBack[since v3.8.4]: used if a more specific group is not defined- its
preferfield replacedoptIn.configStyleArguments, enabled by default
- its
callSite[since v3.8.4]: applies to argument clauses in method calls- replaced section called
newlines.configStyleCallSite
- replaced section called
bracketCallSite[since v3.8.4]: applies to type argument clauses, falls back tocallSiteif not specifieddefnSite[since v3.8.4]: applies to parameter clauses in method or class definitions- replaced section called
newlines.configStyleDefnSite
- replaced section called
bracketDefnSite[since v3.8.4]: applies to type parameter clauses, falls back todefnSiteif not specified
newlines.configStyle.xxxSite.prefer
If true, applies config-style formatting:
newlines.source = fold/unfold: if single-line formatting is impossiblenewlines.source = classic/keep: if preference for config-style is detected in the source:- there's a newline present before the closing delimiter
- if
danglingParentheses.xxxSite = false(andalign.openParenXxxSite = false): no other condition needs to be present (the scala.js rule) - otherwise, there needs to be a newline after the opening delimiter,
or after the
implicitkeyword if present.
Please note that other parameters might also force config-style (see below).
object a {
// keeps single line
def method1(a: Int, b: String): Boolean
// forces config style
def method2(
a: Int,
b: String,
c: String
): Boolean
// preserves config style
def method3(
a: Int,
b: String,
c: String
): Boolean
}
object a {
// keeps single line
def method1(a: Int, b: String): Boolean
// forces config style
def method2(a: Int, b: String, c: String): Boolean
// preserves config style
def method3(
a: Int, b: String, c: String
): Boolean
}
Config for this example:
newlines.configStyle.defnSite.prefer = true
maxColumn=45
Forcing config style
When newlines.configStyle.xxxSite.forceIfOptimized is enabled and
route search optimization
is applicable to a clause, the formatter will force config-style formatting.
By default, takes the same value as
newlines.configStyle.xxxSite.prefer.
object a {
// this is a definition, not a method call
def method(a: String, b: String = null): Boolean
// keeps single line; min offset not satisfied
method(
a,
b
)
// keeps single line; min arg not satisfied
method(SomeVeryVeryVeryVeryLongArgument)
// forces config style
method(
foo,
bar
)
}
object a {
// this is a definition, not a method call
def method(a: String, b: String = null): Boolean
// keeps single line; min offset not satisfied
method(a, b)
// keeps single line; min arg not satisfied
method(SomeVeryVeryVeryVeryLongArgument)
// forces config style
method(foo, bar)
}
Config for this example:
newlines.configStyle.callSite.forceIfOptimized = true
newlines.configStyle.defnSite.forceIfOptimized = false
runner.optimizer.callSite { minSpan = 5, minCount = 2 }
maxColumn = 60
newlines.configStyle.beforeComma
If enabled, will break before comma (instead of after) when using config style.
However, if the runner.dialect supports trailing commas, using
rewrite.trailingCommas is recommended (which is why,
prior to v3.8.4, this parameter was called poorMansTrailingCommasInConfigStyle).
`
Rewrite Rules
To enable a rewrite rule, add it to the config like this
rewrite.rules = [Imports].
AvoidInfix
This rule replaces infix expressions a op b with proper method calls a.op(b).
NB: The rule currently does not support right-associative operators (i.e., those which end in
:) which would have had to be rewritten asb.op(a).
The rule takes the following parameters under rewrite.avoidInfix:
includeFiltersandexcludeFilters, two lists of regular expressions, which determine which operators are eligible for this rewrite- for this rule to be enabled (that is, for the rewrite to be applied), an infix
expression must match
includeFiltersand not matchexcludeFilters - (since 3.8.0) if a regular expression contains
\\., matching will be against not only the infix operator but also its left-hand-side expression (with the non-empty operator part following the last\\.in the pattern) - (before 3.8.0) these two parameters were nested under
rewrite.neverInfix
- for this rule to be enabled (that is, for the rewrite to be applied), an infix
expression must match
excludePlaceholderArg(default:true) will not rewrite infix expressions if the argument is a solo placeholder (_or(_: Type))- this parameter does not control any other cases with the infix argument containing a placeholder character; some of them will never be rewritten as adding parentheses will change their syntactic meaning, and others will be rewritten as usual
- (before 3.8.0 and since 3.4.4) this parameter was named
rewrite.allowInfixPlaceholderArg
- (since 3.8.0)
excludeScalaTestcontrols whether the standard set ofscalatestassert methods is added toexcludeFilters- if unspecified, and
project.layoutdetermines that the file being formatted is not a test file, then these test assert methods will not be excluded
- if unspecified, and
- (since 3.8.4)
excludePostfix, unless set totrueexplicitly, will also apply the rule toTerm.Selecttrees specified without a dot - (since 3.8.4)
excludeMatch, if set tofalseexplicitly and if the dialect enablesallowMatchAsOperator(such as Scala 3), will also apply the rule toTerm.Matchtrees specified without a dot
a.success(b)
a.error(b, c)
a map { x =>
x + 2
}
("o" % "a" % "v").c(D)
(future map { case e: Err =>
0
}).recover(_.toString)
future.recover { case e: Err =>
0
} map (_.toString)
a success b
a error (b, c)
a map { x =>
x + 2
}
"o" % "a" % "v" c(D)
future map {
case e: Err => 0
} recover (_.toString)
future recover {
case e: Err => 0
} map (_.toString)
Config for this example:
rewrite.rules = [AvoidInfix]
rewrite.avoidInfix.excludeFilters."+" = [ "map" ]
_.foo(_)
_.bar(_: Int)
_.baz(_.qux)
_ baz _.qux // cannot be rewritten, not the same as previous line
_ foo _
_ bar (_: Int)
_ baz (_.qux)
_ baz _.qux // cannot be rewritten, not the same as previous line
Config for this example:
rewrite.rules = [AvoidInfix]
rewrite.avoidInfix.excludePlaceholderArg = false
RedundantBraces
Warning. This rewrite can cause non-idempotent formatting, see #1055.
def foo =
List(1, 2, 3).sum
def foo = {
List(1, 2, 3).sum
}
Config for this example:
rewrite.rules = [RedundantBraces]
q"Hello $name"
q"Hello ${name}"
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.stringInterpolation = true
List(1, 2, 3).map(x => x + 1)
List(1, 2, 3).map { x => x + 1 }
Config for this example:
rewrite.rules = [RedundantBraces]
Entire power of RedundantBraces can be accessed with
newlines.afterCurlyLambdaParams=squash. It will try to squash lambda body in
one line and then replace braces with parens:
List(1, 2, 3).map(x => x + 1)
List(1, 2, 3).map { x =>
println("you can't squash me!")
x + 1
}
List(1, 2, 3).map { x =>
x + 1
}
List(1, 2, 3).map { x =>
println("you can't squash me!")
x + 1
}
Config for this example:
rewrite.rules = [RedundantBraces]
newlines.afterCurlyLambdaParams=squash
RedundantBraces: generalExpressions
Default: rewrite.redundantBraces.generalExpressions = true
while (x < 10)
x += 1
str match {
case "a" =>
println("ok")
case _ =>
println("not ok")
}
while (x < 10) {
x += 1
}
str match {
case "a" => {
println("ok")
}
case _ => {
println("not ok")
}
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.generalExpressions = true
RedundantBraces: ifElseExpressions
Default: rewrite.redundantBraces.ifElseExpressions = false
if (a > b)
doSomething()
else
doAnything()
if (a > b) {
doSomething()
} else {
doAnything()
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.ifElseExpressions = true
RedundantBraces: defnBodies
This parameter takes the following values:
none: rewriting of a definition body is disabledall: applies to body of any definition (def,val,macroetc.)noParams: applies to body of any definition which doesn't have parameters (e.g.:val;var; parameterlessdef, without brackets or parentheses)
In v3.3.2, this parameter superseded a boolean methodBodies.
Default: rewrite.redundantBraces.defnBodies = "all"
def f() =
1 + 1
def f() = {
1 + 1
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.defnBodies = true
RedundantBraces: includeUnitMethods
Default: rewrite.redundantBraces.includeUnitMethods = true
Affects only functions with explicitly specified Unit type
def f() =
1 + 1
def x(): Unit = {
println("example")
}
def f() = {
1 + 1
}
def x(): Unit = {
println("example")
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.methodBodies = true
rewrite.redundantBraces.includeUnitMethods = false
RedundantBraces: stringInterpolation
Default: rewrite.redundantBraces.stringInterpolation = false
s"user id is $id"
s"user id is ${id}"
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.stringInterpolation = true
RedundantBraces: oneStatApply
Added in v3.8.4, controls treatment of single-argument apply delimiters, has lower priority than fewer braces, and takes the following values:
parensMaxSpan: if non-negative, converts braces to parentheses in an argument clause if its span is not larger than the specified value (and can be legally used with parentheses)bracesMinSpan: if non-negative, converts parentheses to braces in a single-statement argument clause if its span is strictly larger than the specified value- clearly,
parensMaxSpanmay not exceedbracesMinSpanbut might be the same (for convenience of specifying the pivot point)
- clearly,
parensMaxSpan=0, bracesMinSpan<0is a special combination which replaced an earlier parameterparensForOneLineApply = trueand uses parens when the argument clause ends up formatted on a single line (rather than looking at the argument clause span)
Here, the span is computed a bit differently than for fewer braces or search optimizer, in that it removes not only whitespace but also all punctuation (opening and closing delimiters, commas, semicolons and dots), comments and any optional syntax tokens in scala3 (such as
then,:in coloneol or varargs, end markers etc.), anything that a rewrite rule could modify.The reason is that this happens during the rewrite phase where this or other rules could modify or remove braces, trim trailing commas, add dots and parentheses to an infix, rewrite scala3 code using new syntax, or reformat comments.
Thus, to avoid non-idempotent formatting, we ignore them.
In all cases, redundant delimiters will be rewritten, as before.
# Defaults
rewrite.redundantBraces.oneStatApply.parensMaxSpan = 0
rewrite.redundantBraces.oneStatApply.bracesMinSpan = -1
xs.map(x => x + 1)
xs.map { x => x + 1 }
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.oneStatApply.parensMaxSpan = 0
xs.map(x => // should rewrite this
x + 1
)
xs.map {
x => // should not rewrite this, not a single-stat argument
x + 1
x + 2
}
xs.map {
x => // should not rewrite outer, span too long
x.foo(
bar.baz(qux)
)
}
xs.map { x => // should rewrite this
x + 1
}
xs.map { x => // should not rewrite this, not a single-stat argument
x + 1
x + 2
}
xs.map { x => // should not rewrite outer, span too long
x.foo {
bar.baz(qux)
}
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.oneStatApply.parensMaxSpan = 15
xs.map { x => // should rewrite this
x + 1
}
xs.map( // should not rewrite this, contains an infix with a break before operator
x
+ 1
)
xs.map { // should rewrite this, contains an infix with a break after operator
x +
1
}
// scalafmt: { newlines.afterInfix = some }
xs.map { // should rewrite this, allows formatting infix
x + 1
}
xs.map( // should not rewrite this, span too short
xy
)
xs.map(x => // should rewrite this
x + 1
)
xs.map( // should not rewrite this, contains an infix with a break before operator
x
+ 1
)
xs.map( // should rewrite this, contains an infix with a break after operator
x +
1
)
// scalafmt: { newlines.afterInfix = some }
xs.map( // should rewrite this, allows formatting infix
x
+ 1
)
xs.map( // should not rewrite this, span too short
xy)
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.oneStatApply.bracesMinSpan = 2
RedundantBraces: maxBreaks
This parameter limits the number of line breaks inside the input body. Prior to
v3.3.2, was incorrectly called maxLines.
Default: rewrite.redundantBraces.maxBreaks = 100
def f() =
collection
.map(x => x + 1)
.filter(_ < 10)
.map(_ * 2)
def f() = {
collection
.map(x => x + 1)
.filter(_ < 10)
.map(_ * 2)
.headOption
}
def f() = {
collection
.map(x => x + 1)
.filter(_ < 10)
.map(_ * 2)
}
def f() = {
collection
.map(x => x + 1)
.filter(_ < 10)
.map(_ * 2)
.headOption
}
Config for this example:
rewrite.rules = [RedundantBraces]
rewrite.redundantBraces.maxBreaks = 3
Inserting braces
Warning: this rewrite might cause non-idempotent formatting, formatter might need to be run twice.
This rule cannot be used with
rewrite.scala3.insertEndMarkerMinLinesorrewrite.scala3.removeOptionalBraces.oldSyntaxToo == true.
This rewrite in essence provides the opposite of what RedundantBraces achieves,
and somewhat similar to Scala3's end marker rewrite rules.
The rule is applied after all whitespace decisions had been made and simply attempts to output curly braces around a single-statement block when it spans at least a given number of lines.
The rule is enabled by configuring rewrite.insertBraces:
minLines(default: 0, or disabled): the minimum number of lines to trigger the ruleallBlocks(default: false): compute maximum span of all blocks under the parent expression rather than just the statement to be enclosed in curly braces- this could be used to have consistent application of curly braces in expressions
with multiple sub-expressions (conditions or blocks), such as
if-else,try-finally,for-yield,do-whileetc.
- this could be used to have consistent application of curly braces in expressions
with multiple sub-expressions (conditions or blocks), such as
Here are some limitations:
- the rule might occasionally lead to non-idempotent formatting (that is, applying
the formatter a second time would produce a different result); some examples are:
- adding braces might overflow a line
- adding braces might lead to different indentation of infix expressions
- the rule will not be applied:
- unless the single statement is preceded by a newline; doing so would definitely lead to non-idempotent formatting
- if the statement is an infix expression which is not enclosed in parentheses and has a line break before an operator
- if the code uses Scala3 syntax with significant indentation
RedundantParens
for {
a <- b
if a.nonEmpty
} yield a
val z = insertData *> readDatabase(id)
for {
a <- b
if (a.nonEmpty)
} yield a
val z = (insertData *> readDatabase(id))
Config for this example:
rewrite.rules = [RedundantParens]
RedundantParens: infixSide
Since v3.5.4.
Parameter rewrite.redundantParens.infixSide controls how the rule applies
to expressions which are part of an outer infix expression (either left- or
right-hand side). Can take the following values:
null(default): rewrites only simple expressions (literals or identifiers)
// null+
foo & 0 // literal
foo & bar // identifier
// some+
foo & (bar.baz) // non-infix
foo & (bar + baz) // very high precedence infix
foo or (bar < baz) // non-symbolic outer op, medium precedence infix
// null+
foo & (0) // literal
foo & (bar) // identifier
// some+
foo & (bar.baz) // non-infix
foo & (bar + baz) // very high precedence infix
foo or (bar < baz) // non-symbolic outer op, medium precedence infix
Config for this example:
rewrite.rules = [RedundantParens]
rewrite.redundantParens.infixSide = null
some: additionally, rewrites- all non-infix sides
- very-high-precedence nested infix sides
- medium-precedence nested infix sides if the outer infix operator is non-symbolic
// some+
foo & bar.baz // non-infix
foo & bar + baz // very high precedence infix
foo or bar < baz // non-symbolic outer op, medium precedence infix
// many+
foo || (bar == baz) // high precedence infix
foo or (bar || baz) // non-symbolic outer op; low precedence infix
foo |: (bar |: baz) // identical op: non-symbolic; right infix, right assoc
(foo :| bar) :| baz // identical op: symbolic; left infix, left assoc
(foo or bar) or baz // identical op: non-symbolic; left infix, left assoc
// some+
foo & (bar.baz) // non-infix
foo & (bar + baz) // very high precedence infix
foo or (bar < baz) // non-symbolic outer op, medium precedence infix
// many+
foo || (bar == baz) // high precedence infix
foo or (bar || baz) // non-symbolic outer op; low precedence infix
foo |: (bar |: baz) // identical op: non-symbolic; right infix, right assoc
(foo :| bar) :| baz // identical op: symbolic; left infix, left assoc
(foo or bar) or baz // identical op: non-symbolic; left infix, left assoc
Config for this example:
rewrite.rules = [RedundantParens]
rewrite.redundantParens.infixSide = some
many: additionally, rewrites- high-precedence nested infix sides
- nested infix sides when the operator is identical to the outer infix (and associativity allows)
- any symbolic nested infix sides if the outer infix operator is non-symbolic
// many+
foo || bar == baz // high precedence infix
foo or bar || baz // non-symbolic outer op; low precedence infix
foo |: bar |: baz // identical op: non-symbolic; right infix, right assoc
foo :| bar :| baz // identical op: symbolic; left infix, left assoc
foo or bar or baz // identical op: non-symbolic; left infix, left assoc
// all
foo || (bar && baz) // low precedence infix
// many+
foo || (bar == baz) // high precedence infix
foo or (bar || baz) // non-symbolic outer op; low precedence infix
foo |: (bar |: baz) // identical op: non-symbolic; right infix, right assoc
(foo :| bar) :| baz // identical op: symbolic; left infix, left assoc
(foo or bar) or baz // identical op: non-symbolic; left infix, left assoc
// all
foo || (bar && baz) // low precedence infix
Config for this example:
rewrite.rules = [RedundantParens]
rewrite.redundantParens.infixSide = many
all: rewrites all expressions within an infix
// all
foo || bar && baz // low precedence infix
// all
foo || (bar && baz) // low precedence infix
Config for this example:
rewrite.rules = [RedundantParens]
rewrite.redundantParens.infixSide = all
SortModifiers
Modifiers are sorted based on the given order. Affects modifiers of the following definitions: trait, class, object, type, and val+var, both as fields and class parameters.
implicit final private lazy val x = 42
implicit final private lazy val y = 42
class Test(implicit
final private val i1: Int,
final private val i2: String
)
sealed protected[X] trait ADT
final private case object A1 extends ADT
final private case class A2(a: Int)
extends ADT
final lazy private implicit val x = 42
lazy final implicit private val y = 42
class Test(
implicit
final private val i1: Int,
private final val i2: String
)
sealed protected[X] trait ADT
final private case object A1 extends ADT
private final case class A2(a: Int)
extends ADT
Config for this example:
rewrite.rules = [SortModifiers]
If you prefer to use the order based on the one partially specified in the Scala Style Guide, you can use the corresponding preset (since v3.8.1):
private implicit final lazy val x = 42
private implicit final lazy val y = 42
class Test(implicit
private final val i1: Int,
private final val i2: String
)
protected[X] sealed trait ADT
private final case object A1 extends ADT
private final case class A2(a: Int)
extends ADT
final lazy private implicit val x = 42
lazy final implicit private val y = 42
class Test(
implicit
final private val i1: Int,
private final val i2: String
)
sealed protected[X] trait ADT
final private case object A1 extends ADT
private final case class A2(a: Int)
extends ADT
Config for this example:
rewrite.rules = [SortModifiers]
rewrite.sortModifiers.preset = styleGuide
If you choose a custom sort order, you can specify some or all modifiers in the desired order:
implicit final override val x = 2
override implicit final val x = 2
Config for this example:
rewrite.rules = [SortModifiers]
rewrite.sortModifiers.order = [
"implicit", "final", "sealed", "abstract",
"override", "private", "protected", "lazy"
]
Hint: since some modifiers are mutually exclusive, you might want to order them next to each other.
If you fail to include some modifiers, they will be moved to the front, before modifiers explicitly configured, while preserving their relative order to each other.
final implicit override val x = 2
override implicit final val x = 2
Config for this example:
rewrite.rules = [SortModifiers]
rewrite.sortModifiers.order = [
"implicit", "override"
]
The following modifiers are recognized currently:
overrideprivate,protectedimplicitfinal,sealed,abstracterased,lazy,open,transparent,inline,infix,opaque
PreferCurlyFors
Replaces parentheses into curly braces in for comprehensions that contain multiple enumerator generators.
for {
a <- as
b <- bs if b > 2
} yield (a, b)
for(a <- as; b <- bs if b > 2)
yield (a, b)
Config for this example:
rewrite.rules = [PreferCurlyFors]
This rule accepts the following settings:
rewrite.preferCurlyFors.removeTrailingSemicolonsOnly(default:false):- if
false(default), replaces all semicolons with a newline - if
true, keeps semicolons unless followed by a newline or single-line comment
- if
Imports
This rule applies to import or, in Scala 3, also export statements found
at the top level (source, package or class/object/trait level).
The logic also moves comments attached to each import statement or selector, as follows:
- all consecutive standalone comments before (no blank lines and not following some other token)
- the comment right after (possibly after a comma), on the same line
- only single-line (
//) comments are supported; multiline (/*) comments will not be moved and in many cases will likely be removed instead
Since v3.0.0.
Imports: expand
This parameter will attempt to create a separate line for each selector
within a {...}. It replaces the deprecated rule ExpandImportSelectors.
Default: rewrite.imports.expand = false
import a.b
import a.c
import h.k
import h.l
import d.e.f
import d.e.g
import a.{foo => bar, zzzz => _, _}
import a.{
b,
c
}, h.{
k, l
}
import d.e.{f, g}
import a.{
foo => bar,
zzzz => _,
_
}
Config for this example:
rewrite.rules = [Imports]
rewrite.imports.expand = true
Imports: sort = none
This default option causes no sorting.
Imports: sort = original
Replaces the deprecated rule SortImports.
The imports are sorted by the groups: symbols, lower-case, upper-case.
import foo.{bar, sand, Random, Zilch}
import foo.{Zilch, bar, Random, sand}
Config for this example:
rewrite.rules = [Imports]
rewrite.imports.sort = original
Imports: sort = ascii
Replaces the deprecated rule AsciiSortImports.
The imports are sorted by their Ascii codes.
import foo.{Random, `symbol`, bar, ~>}
import foo.{~>, `symbol`, bar, Random}
Config for this example:
rewrite.rules = [Imports]
rewrite.imports.sort = ascii
Imports: sort = scalastyle
- Selectors are sorted in a case-insensitive manner (ascii on lowercase), except by first character as follows: non-wildcard symbols, lowercase, uppercase, wildcard.
- If grouping, import statements are also sorted using case-insensitive order, except by first character in every dot-separated label as follows: symbols (including wildcard), uppercase, lowercase.
import foo._
import foo.`qux`.{`symbol`, ~>, bar, Random}
import foo.Baz.{bar => xyz, _}
import foo.bar.{`symbol`, ~>, bar, Random}
import foo.bar.{Random, bar, ~>, `symbol`}
import foo.Baz.{bar => xyz, _}
import foo.`qux`.{Random, bar, ~>, `symbol`}
import foo._
Config for this example:
maxColumn = 50
rewrite.rules = [Imports]
rewrite.imports.sort = scalastyle
rewrite.imports.groups = [["foo\\..*"]]
Imports: groups
Keep in mind that this functionality should be used very carefully if hierarchical (relative) imports are allowed in your codebase. Groups should only refer to typical top-level domains such as
java,org,comorscala, and sorting should be disabled.The safest way to handle this case is by using
scalafixwith a semantic rule likeOrganizeImports. However, on a large codebase, the overhead of using semanticscalafixrules might be substantial.
This rule will separate all import statements into groups. If sorting is
enabled (i.e., not none), imports will also be sorted within each group.
The rule accepts the following parameters:
rewrite.imports.groups: defines several sets of regular expressions- each set defines a single group, and the groups are output in the order they are configured
- imports not matching any of the regexes will form their own group at the end
- regular expressions are applied to the entire parent domain of the import statement, up to and including the final dot
- the longest patterns are applied first
rewrite.imports.contiguousGroups(since v3.0.2):- if
only(default), only consecutive import statements will be grouped - if
no, grouping will happen on all imports within the same container (source, package, template etc.)
- if
import foo.Baz.{bar => xyz, _}
import foo._
import foo.`qux`.{
Random,
`symbol`,
bar,
~>
}
import bar._
import bar.`qux`.{
Random,
`symbol`,
bar,
~>
}
import bar.bar.{
Random,
`symbol`,
bar,
~>
}
import baz.Baz.{bar => xyz, _}
import baz._
import baz.bar.{
Random,
`symbol`,
bar,
~>
}
import qux.Baz.{bar => xyz, _}
import qux.`qux`.{
Random,
`symbol`,
bar,
~>
}
import qux.bar.{
Random,
`symbol`,
bar,
~>
}
import bar.bar.{Random, bar, ~>, `symbol`}
import baz.Baz.{bar => xyz, _}
import qux.`qux`.{Random, bar, ~>, `symbol`}
import foo._
import baz.bar.{Random, bar, ~>, `symbol`}
import qux.Baz.{bar => xyz, _}
import foo.`qux`.{Random, bar, ~>, `symbol`}
import bar._
import qux.bar.{Random, bar, ~>, `symbol`}
import foo.Baz.{bar => xyz, _}
import bar.`qux`.{Random, bar, ~>, `symbol`}
import baz._
Config for this example:
rewrite.rules = [Imports]
rewrite.imports.sort = ascii
rewrite.imports.groups = [
["foo\\..*"],
["bar\\..*", "baz\\..*"]
]
Trailing commas
See SIP
The rule handles how trailing commas are treated in case of a dangling closing delimiter (parenthesis or bracket for definitions or invocations, brace for import statements only).
Regardless of the setting, trailing commas are always removed if the closing delimiter is not dangling (i.e., follows the final argument without a line break).
This logic is not triggered via the
rewrite.rulesparameter, but by setting parameters within therewrite.trailingCommassection (since v3.0.5; prior to that there was a single top-leveltrailingCommasparameter).
Default: rewrite.trailingCommas.style = "never"
Trailing commas: never
Makes sure there are no trailing commas:
import a.{b, c}
def method1(
a: Int,
b: Long
) = {}
def method2(
a: Int,
b: Long*
) = {}
def method3(
a: Int
) = {}
method1(
a,
b
)
method2(
a,
b: _*
)
method3(
a
)
import a.{
b,
c,
}
def method1(
a: Int,
b: Long,
) = {}
def method2(
a: Int,
b: Long*,
) = {}
def method3(
a: Int,
) = {}
method1(
a,
b,
)
method2(
a,
b: _*,
)
method3(
a,
)
Config for this example:
rewrite.trailingCommas.style = never
Trailing commas: keep
Keeps any trailing commas:
import a.{b, c}
def method1(
a: Int,
b: Long,
) = {}
def method2(
a: Int,
b: Long*
) = {}
def method3(
a: Int,
) = {}
method1(
a,
b
)
method2(
a,
b: _*,
)
method3(
a,
)
import a.{
b,
c,
}
def method1(
a: Int,
b: Long,
) = {}
def method2(
a: Int,
b: Long*
) = {}
def method3(
a: Int,
) = {}
method1(
a,
b
)
method2(
a,
b: _*,
)
method3(
a,
)
Config for this example:
rewrite.trailingCommas.style = keep
Trailing commas: always
Makes sure there are trailing commas:
import a.{b, c}
def method1(
a: Int,
b: Long,
) = {}
def method2(
a: Int,
b: Long*,
) = {}
def method3(
a: Int,
) = {}
method1(
a,
b,
)
method2(
a,
b: _*,
)
method3(
a,
)
import a.{
b,
c
}
def method1(
a: Int,
b: Long
) = {}
def method2(
a: Int,
b: Long*
) = {}
def method3(
a: Int
) = {}
method1(
a,
b
)
method2(
a,
b: _*
)
method3(
a
)
Config for this example:
rewrite.trailingCommas.style = always
Trailing commas: multiple
Since v2.5.0.
Makes sure there are trailing commas for multiple-argument expressions only, except when the last argument is repeated:
import a.{b, c}
def method1(
a: Int,
b: Long,
) = {}
def method2(
a: Int,
b: Long*
) = {}
def method3(
a: Int
) = {}
method1(
a,
b,
)
method2(
a,
b: _*
)
method3(
a
)
import a.{
b,
c
}
def method1(
a: Int,
b: Long
) = {}
def method2(
a: Int,
b: Long*,
) = {}
def method3(
a: Int,
) = {}
method1(
a,
b
)
method2(
a,
b: _*,
)
method3(
a,
)
Config for this example:
rewrite.trailingCommas.style = multiple
rewrite.trailingCommas.allowFolding
This parameter controls whether the trailing comma must be maintained (except
for never), or the code can be folded to avoid a dangling closing delimiter
which is required by Scala after a trailing comma.
Since v3.0.5
Default: rewrite.trailingCommas.allowFolding = true
If set to false, the trailing comma will always be forced.
rewrite.tokens
Prior to v3.8.4, was called
rewriteTokens.
Map of tokens to rewrite. For example, Map("⇒" -> "=>") will rewrite unicode arrows to regular ascii arrows.
Default: rewrite.tokens = {}
val tuple = "a" -> 1
val lambda = (x: Int) => x + 1
for {
a <- Option(1)
b <- Option(2)
} yield a + b
val tuple = "a" → 1
val lambda = (x: Int) ⇒ x + 1
for {
a ← Option(1)
b ← Option(2)
} yield a + b
Config for this example:
rewrite.tokens = {
"⇒": "=>"
"→": "->"
"←": "<-"
}
Scala3 rewrites
This section describes rules which are applied if the appropriate dialect (e.g.,
runner.dialect = scala3) is selected.
This logic is not triggered via the
rewrite.rulesparameter, but by setting parameters underrewrite.scala3subsection.
rewrite.scala3.convertToNewSyntax
If this flag is enabled, the following new syntax will be applied (also,
since 3.8.0, if an appropriate flag under rewrite.scala.newSyntax is not
set to false, see below):
- control syntax
- if dialect sets
allowSignificantIndentation(any scala3 dialect) and...newSyntax.controlis setif (...)toif ... thenwhile (...)towhile ... dofor (...)tofor ... do(orfor (...) yieldtofor ... yield)
- if dialect sets
- vararg splices
- vararg
: _*or@ _*to*if dialect setsallowPostfixStarVarargSplices(any scala3, or scala2xxSource3) and...newSyntax.deprecatedis set
- vararg
- imports
- import wildcard
_to*if dialect setsallowStarWildcardImport(any scala3, or scala2xxSource3) and...newSyntax.deprecatedis set - import rename
=>toasif dialect setsallowAsForImportRename(any scala3, or scala2xxSource3) and...newSyntax.deprecatedis set
- import wildcard
- wildcards
- type wildcard
_to?if dialect setsallowQuestionMarkAsTypeWildcard(scala212 and later) and...newSyntax.deprecatedis set - anonymous type param
*to_if dialect setsallowUnderscoreAsTypePlaceholder(scala3Future only) and...newSyntax.deprecatedis set
- type wildcard
NB: You could control these rules individually by overriding dialect properties.
rewrite.scala3.removeOptionalBraces
If this section is enabled, optional braces will be removed and significant indentation applied.
Default: rewrite.scala3.removeOptionalBraces = {"enabled": false, "fewerBracesMinSpan": 2, "fewerBracesMaxSpan": 0, "fewerBracesParensToo": false, "oldSyntaxToo": false}
The section contains the following settings (available since v3.8.1):
enabled:- if
false, disables any rewriting, regardless of other flags in this section - if
true, enables rewriting- applies to expressions using the new control syntax
(or those which would rewritten to new syntax if
rewrite.scala3.convertToNewSyntaxis set) - other flags below might extend rewrites to other cases
- applies to expressions using the new control syntax
(or those which would rewritten to new syntax if
- if
oldSyntaxToo- if
true, applies also to expressions using deprecated syntax
- if
- (since v3.8.1)
fewerBracesMinSpanandfewerBracesMaxSpan- will apply the rewrite to last curried single-argument group if
it is enclosed in curly braces (or would be rewritten to curly
braces by the
RedundantBracesrule) - will only apply the rewrite if the cumulative span of all visible (non-whitespace) tokens within the argument is between the two values
- this rule is disabled if
fewerBracesMaxSpan == 0
- will apply the rewrite to last curried single-argument group if
it is enclosed in curly braces (or would be rewritten to curly
braces by the
- (since v3.8.4)
fewerBracesParensToo- will apply the rule just above to an argument in parentheses as well,
if the one of following is also satisfied:
newlines.afterInfixis NOTkeep; or- current dialect supports
allowInfixOperatorAfterNL
- will apply the rule just above to an argument in parentheses as well,
if the one of following is also satisfied:
Prior to v3.8.1, rewrite.scala3.removeOptionalBraces was a flag which
took three possible values (with their equivalent current settings shown):
no:enabled = falseyes:enabled = trueoldSyntaxToo:enabled = trueandoldSyntaxToo = true
rewrite.scala3.insertEndMarkerMinLines
If this flag is set to a positive value, when an expression containing an optional braces region spans at least as many lines and isn't followed by an end marker, one will be inserted.
We will not insert end markers if the statement is not part of a template body, or a multi-stat block. Doing so might turn a single-stat expression (which doesn't require significant indentation handling) into a multi-stat block.
rewrite.scala3.removeEndMarkerMaxLines
If this flag is set to a positive value, when an expression containing an optional braces region spans at most as many lines and is followed by a standalone end marker (i.e., no other tokens on that line, including comments), the line containing the end marker will be deleted.
We will not remove end markers if
- the statement is not part of a template body, or a block with at least 3 statements. Doing so might turn a multi-stat expression (which requires significant indentation handling) into a single-stat.
- there are comments before the end marker, as without the end marker they would be treated as outside of the optional-braces region.
rewrite.scala3.countEndMarkerLines
Since v3.0.6.
This flag dictates which part of the expression terminated by the end marker
is used to calculate the span for the purposes of applying
insertEndMarkerMinLines and
removeEndMarkerMaxLines.
all(default): the entire expressionlastBlockOnly: only the last block with significant indentation relative to the start of the said expression (as a replacement for the closing curly brace which would have been used otherwise); for instance:- in case of a class, this would be the body of the class
- but for an if-else, this would be just the
elsepart
Vertical Multiline
Since: v1.6.0.
If enabled this formats methods such that parameters are on their own line
indented by indent.defnSite.
Separation between parameter groups are indented by two spaces less than
indent.defnSite. The return type is on its own line at the end.
This formatting is only triggered if the method definition exceeds the maxColumn value in width or if the number of arguments to the method exceeds the
verticalMultiline.arityThreshold.
verticalMultiline.arityThreshold
Default: verticalMultiline.arityThreshold = 100
case class Foo(x: String)
case class Bar(
x: String,
y: String)
object A {
def foo(
x: String,
y: String
)
def hello(
how: String
)(are: String
)(you: String
) = how + are + you
}
case class Foo(x: String)
case class Bar(x: String, y: String)
object A {
def foo(x: String, y: String)
def hello(how: String)(are: String)(you: String) = how + are + you
}
Config for this example:
verticalMultiline.atDefnSite = true
verticalMultiline.arityThreshold = 2
verticalMultiline.newlineAfterOpenParen
Default: verticalMultiline.newlineAfterOpenParen = false
def other(
a: String,
b: String
)(
c: String,
d: String
) = a + b + c
def other(a: String, b: String)(c: String, d: String) = a + b + c
Config for this example:
indent.defnSite = 2
verticalMultiline.atDefnSite = true
verticalMultiline.arityThreshold = 2
verticalMultiline.newlineAfterOpenParen = true
verticalMultiline.excludeDanglingParens
This parameter has been removed in 3.4.0, please use danglingParentheses.exclude.
def other(
a: String,
b: String
)(
c: String,
d: String) = a + b + c
other(a, b)(c, d)
def other(a: String, b: String)(c: String, d: String) = a + b + c
other(a, b)(c, d)
Config for this example:
indent.defnSite = 2
danglingParentheses.exclude = [def]
verticalMultiline.atDefnSite = true
verticalMultiline.arityThreshold = 2
verticalMultiline.newlineAfterOpenParen = true
Vertical multiline with implicit parameter lists
Also see the general section on implicit parameter lists.
Before only
def format(
code: String,
age: Int
)(
implicit ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
verticalMultiline.atDefnSite = true
newlines.implicitParamListModifierForce = [before]
After only
def format(
code: String,
age: Int
)(implicit
ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
verticalMultiline.atDefnSite = true
newlines.implicitParamListModifierForce = [after]
Before and after
def format(
code: String,
age: Int
)(
implicit
ev: Parser,
c: Context
): String
def format(code: String, age: Int)(implicit ev: Parser, c: Context): String
Config for this example:
maxColumn = 60
verticalMultiline.atDefnSite = true
newlines.implicitParamListModifierForce = [before,after]
Comment processing
comments.wrap
Since v2.6.0.
Allows wrapping comments exceeding maxColumn.
Default: comments.wrap = "no"
comments.wrap = standalone
A standalone comment is one which is surrounded by line breaks.
/* long multiline
* comment */
// long singleline comment
val a = 1 // short
val b =
2 // long singleline comment
/* long multiline comment */
// long singleline comment
val a = 1 // short
val b = 2 // long singleline comment
Config for this example:
maxColumn = 20
comments.wrap = standalone
comments.wrap = trailing
A trailing comment is one which is followed by a line break.
/* long multiline
* comment */
// long singleline comment
val a = 1 // short
val b =
2 /* long
* singleline
* comment */
/* long multiline comment */
// long singleline comment
val a = 1 // short
val b = 2 // long singleline comment
Config for this example:
maxColumn = 20
comments.wrap = trailing
comments.wrapStandaloneSlcAsSlc
Since v2.6.0.
This parameter allows formatting a standalone single-line comment (i.e., //)
to be wrapped using the same type, not a multi-line comment (/* ... */).
Default: comments.wrapStandaloneSlcAsSlc = false
// long singleline comment
val b =
2 /* long
* singleline
* comment */
// long singleline comment
val b = 2 // long singleline comment
Config for this example:
maxColumn = 20
comments.wrap = trailing
comments.wrapStandaloneSlcAsSlc = true
comments.wrapSingleLineMlcAsSlc
Since v3.3.1.
If comment wrapping is enabled (comments.wrap != no), this parameter
allows formatting a trailing or standalone multi-line comment (i.e., /* ... */) as a
single-line comment (//) if it occupies a single line.
Default: comments.wrapSingleLineMlcAsSlc = false
// standalone multi-line comment
val b = 2 /* mlc */ // trailing mlc
/* standalone multi-line comment */
val b = 2 /* mlc */ /* trailing mlc */
Config for this example:
maxColumn = 50
comments.wrap = trailing
comments.wrapSingleLineMlcAsSlc = true
docstrings.style
Since v2.6.0.
Default: docstrings.style = "SpaceAsterisk"
docstrings.style = keep
Prohibits formatting of docstrings. All other docstrings parameters are
ignored.
Since v3.0.0.
/** do not touch
* this style
* keep the text as-is
*/
/** do not touch
* this style
* keep the text as-is
*/
Config for this example:
docstrings.style = keep
docstrings.style = Asterisk
This variant used to be called JavaDoc.
/**
* Skip first line, format intermediate
* lines with an asterisk below the
* first asterisk of the first line (aka
* JavaDoc). Since v3.8.4,
* `blankFirstLine = fold` takes
* precedence.
*/
/** Skip first line, format intermediate lines with an asterisk
* below the first asterisk of the first line (aka JavaDoc).
* Since v3.8.4, `blankFirstLine = fold` takes precedence.
*/
Config for this example:
docstrings.style = Asterisk
docstrings.style = SpaceAsterisk
This variant used to be called ScalaDoc.
/** Format intermediate lines with a
* space and an asterisk, both below
* the two asterisks of the first line
*/
/** Format intermediate lines with a space and an asterisk,
* both below the two asterisks of the first line
*/
Config for this example:
docstrings.style = SpaceAsterisk
docstrings.style = AsteriskSpace
/** Format intermediate lines with an
* asterisk and a space, both below the
* two asterisks of the first line
*/
/** Format intermediate lines with an asterisk and a space,
* both below the two asterisks of the first line
*/
Config for this example:
docstrings.style = AsteriskSpace
docstrings.removeEmpty
If set, will cause empty docstrings to be removed.
Since v3.0.4.
Default: docstrings.removeEmpty = false
val a = 1
/** */
/**
*
*/
/** */ /** */
val a = 1
Config for this example:
docstrings.removeEmpty = true
docstrings.oneline
Since v2.6.0. Ignored for
docstrings.style = keep.
Default: docstrings.oneline = "keep"
docstrings.oneline = fold
/** Scaladoc oneline */
/** Scaladoc multiline */
val a = 1
/** Scaladoc oneline */
/**
* Scaladoc multiline
*/
val a = 1
Config for this example:
docstrings.style = Asterisk
docstrings.oneline = fold
docstrings.oneline = unfold
/**
* Scaladoc oneline
*/
/**
* Scaladoc multiline
*/
val a = 1
/** Scaladoc oneline */
/**
* Scaladoc multiline
*/
val a = 1
Config for this example:
docstrings.style = Asterisk
docstrings.oneline = unfold
docstrings.oneline = keep
/** Scaladoc oneline */
/**
* Scaladoc multiline
*/
val a = 1
/** Scaladoc oneline */
/**
* Scaladoc multiline
*/
val a = 1
Config for this example:
docstrings.style = Asterisk
docstrings.oneline = keep
docstrings.wrap
Will parse scaladoc comments and reformat them. Takes the following values:
keep: preserves scaladoc comments as-is and will not reformat them (replacednoin v3.8.2)fold: will use a more compact, horizontal formatting (added in v3.8.2)unfold: will use a more expanded, vertical formatting (replacedyesin v3.8.2)
This functionality is generally limited to standard scaladoc elements and might lead to undesirable results in corner cases; for instance, the scaladoc parser doesn't have proper support of embedded HTML.
However, tables are supported.
Since v2.6.0. Ignored for
docstrings.style = keep.
Default: docstrings.wrap = "unfold"
/** @param d
* the Double to square,
* meaning multiply by
* itself
* @return
* the result of squaring d
*
* Thus
* - if [[d]] represents a
* negative value:
* a. the result will be
* positive
* a. the value will be
* {{{d * d}}}
* a. it will be the same
* as for `-d`
* - however, if [[d]] is
* positive
* - the value will still
* be {{{d * d}}}
* - i.e., the same as
* {{{(-d) * (-d)}}}
*
* In other words:
* {{{
* res = d * d
* = (-d) * (-d)
* }}}
*/
def pow2(d: Double): Double
/**
* @param d the Double to square, meaning multiply by itself
* @return the result of squaring d
*
* Thus
* - if [[d]] represents a negative value:
* a. the result will be positive
* a. the value will be {{{d * d}}}
* a. it will be the same as for `-d`
* - however, if [[d]] is positive
* - the value will still be {{{d * d}}}
* - i.e., the same as {{{(-d) * (-d)}}}
*
* In other words:
* {{{
* res = d * d
* = (-d) * (-d) }}}
*/
def pow2(d: Double): Double
Config for this example:
docstrings.wrap = yes
maxColumn = 30
docstrings.wrapMaxColumn
If wrapping (or applying oneline), allows specifying a different value than
the default maxColumn.
docstrings.blankFirstLine
Controls whether to force the first line to be blank in a multiline docstring.
Some values of docstrings.style might take precedence.
Takes the following values:
keep: preserves the first line as-isfold: will ensure there's no blank first line (default; replacednoin v3.8.2)unfold: will enforce a blank first line (replacedyesin v3.8.2)
Since v2.7.5.
- Ignored for
docstrings.style = keepordocstrings.wrap = no.- [since v3.8.4] For
docstrings.style = Asterisk, onlyfoldchanges default behaviour.
/** Scaladoc oneline */
/** Scaladoc multiline1
*/
/** Scaladoc multiline2
*/
val a = 1
/** Scaladoc oneline */
/** Scaladoc multiline1
*/
/**
* Scaladoc multiline2
*/
val a = 1
Config for this example:
# do not force a blank first line
docstrings.blankFirstLine = fold
docstrings.style = SpaceAsterisk
maxColumn = 30
/** Scaladoc oneline */
/**
* Scaladoc multiline1
*/
/**
* Scaladoc multiline2
*/
val a = 1
/** Scaladoc oneline */
/** Scaladoc multiline1
*/
/**
* Scaladoc multiline2
*/
val a = 1
Config for this example:
# force a blank first line
docstrings.blankFirstLine = unfold
docstrings.style = SpaceAsterisk
maxColumn = 30
/** Scaladoc oneline */
/** Scaladoc multiline1
*/
/** Scaladoc multiline2
*/
val a = 1
/** Scaladoc oneline */
/** Scaladoc multiline1
*/
/**
* Scaladoc multiline2
*/
val a = 1
Config for this example:
# preserve a blank first line
docstrings.blankFirstLine = keep
docstrings.style = SpaceAsterisk
maxColumn = 30
docstrings.forceBlankLineBefore
If true (default), always insert a blank line before docstrings. If false, preserves blank line only if one exists before.
Since v3.4.0. Replaced deprecated
optIn.forceBlankLineBeforeDocstring.
object Stuff {
/** Some function */
def hello = ()
}
object Stuff {
/** Some function */
def hello = ()
}
Config for this example:
docstrings.forceBlankLineBefore = true
object Stuff {
/** Some function */
def hello = ()
}
object Stuff {
/** Some function */
def hello = ()
}
Config for this example:
docstrings.forceBlankLineBefore = false
Disabling or customizing formatting
Search state exploded
If this exception occurs, you can try increasing optimizer limits, globally or using any of the options further in this section.
For code block
There is a possibility to override scalafmt config for a specific code with
// scalafmt: {} comment:
// scalafmt: { align.preset = most, danglingParentheses.preset = false }
libraryDependencies ++= Seq(
"org.scalameta" %% "scalameta" % scalametaV,
"org.scalacheck" %% "scalacheck" % scalacheckV)
// scalafmt: { align.preset = some, danglingParentheses.preset = true } (back to defaults)
libraryDependencies ++= Seq(
"org.scalameta" %% "scalameta" % scalametaV,
"org.scalacheck" %% "scalacheck" % scalacheckV
)
// scalafmt: { align.preset = most, danglingParentheses.preset = false }
libraryDependencies ++= Seq(
"org.scalameta" %% "scalameta" % scalametaV,
"org.scalacheck" %% "scalacheck" % scalacheckV)
// scalafmt: { align.preset = some, danglingParentheses.preset = true } (back to defaults)
libraryDependencies ++= Seq(
"org.scalameta" %% "scalameta" % scalametaV,
"org.scalacheck" %% "scalacheck" % scalacheckV)
// format: off
Disable formatting for specific regions of code by wrapping them in
// format: off blocks:
// format: off
val identity = Array(1, 0, 0,
0, 1, 0,
0, 0, 1)
// format: on
// format: off
val identity = Array(1, 0, 0,
0, 1, 0,
0, 0, 1)
// format: on
Since v3.8.4, these format on-off tags can be configured:
formatOff: tags to turn formatting offformatOn: tags to turn formatting back on
# Defaults
formatOff = ["@formatter:off", "format: off"]
formatOn = ["@formatter:on", "format: on"]
Project
Configure which source files should be formatted in this project.
project.git
If this boolean flag is set, only format files tracked by git.
project.include/exclude
Since v3.0.0.
# Defaults
project.includePaths = ["glob:**.scala", "glob:**.sbt", "glob:**.sc", "glob:**.mill"]
project.excludePaths = []
Allows specifying
PathMatcher
selection patterns to identify further which files are to be formatted (explicit glob: or
regex: prefixes are required; keep in mind that PathMatcher patterns must match the entire
path).
For instance,
project {
includePaths = [
"glob:**.scala",
"regex:.*\\.sc"
]
excludePaths = [
"glob:**/src/test/scala/**.scala"
]
}
project.layout
Allows specifying a project structure naming convention which can be used to
select an appropriate dialect for cross-building if one is explicitly selected
via fileOverride. By default, it's disabled (null).
Currently, the following options are supported:
- (since v3.2.0)
StandardConvention: this is the usual naming convention putting scala source code undersrc/main/scalaorsrc/test/scala, with alternate cross-build dialects insrc/main/scala-2.13
If this parameter is set, some supported dialects will be determined automatically;
if the detected dialect is compatible with the overall one (runner.dialect),
no change will be applied.
Currently, supports scala binary versions 2.10-2.13 as well as 3; also, if the version
is major scala 2 (i.e., scala-2), will select the scala 2.13 dialect.
fileOverride
Since v2.5.0.
- This parameter does not modify which files are formatted.
- The match pattern will be applied to the entire absolute, canonical file name; it is not a suffix or a substring match.
Allows specifying an additional subset of parameters for each file matching a
PathMatcher
pattern (e.g., a glob or a regex):
fileOverride {
"<PathMatcher pattern>" { # must match the entire filename
<overridden parameters>
}
}
For instance,
align.preset = none
fileOverride {
"glob:**.sbt" {
align.preset = most
}
"glob:**/src/test/scala/**.scala" {
maxColumn = 120
binPack.callSite = true
}
}
uses align.preset=none for all files except .sbt for which
align.preset=most will apply. It will also use different parameters for test
suites.
File names will be matched against the patterns in the order in which they are specified in the configuration file, in case multiple patterns match a given file.
The parameter also allows the following shortcuts:
- (since v3.2.0) setting only the dialect:
fileOverride { "glob:**/*.sbt" = sbt1 }
- (since v3.2.0) setting based on the file extension:
fileOverride { ".sbt" { runner.dialect = sbt1 } }- this is simply a shortcut for
glob:**.ext
- (since v3.2.0) setting based on the language:
fileOverride { "lang:scala-2" = scala213 }- requires project.layout (sets dialect for minor versions)
- these patterns will be matched last
Spaces
spaces.beforeContextBoundColon
Default: spaces.beforeContextBoundColon = "Never"
def method[A: Bound]: B
def method[A: Bound]: B
def method[A: Bound: Bound2]: B
def method[A: Bound]: B
def method[A : Bound]: B
def method[A: Bound: Bound2]: B
Config for this example:
spaces.beforeContextBoundColon=Never
def method[A : Bound]: B
def method[A : Bound]: B
def method[A : Bound : Bound2]: B
def method[A: Bound]: B
def method[A : Bound]: B
def method[A: Bound: Bound2]: B
Config for this example:
spaces.beforeContextBoundColon=Always
def method[A: Bound]: B
def method[A: Bound]: B
def method[A : Bound : Bound2]: B
def method[A: Bound]: B
def method[A : Bound]: B
def method[A: Bound: Bound2]: B
Config for this example:
spaces.beforeContextBoundColon=IfMultipleBounds
spaces.inImportCurlyBraces
Default: spaces.inImportCurlyBraces = false
import a.b.{ c, d }
import a.b.{c, d}
Config for this example:
spaces.inImportCurlyBraces=true
spaces.inInterpolatedStringCurlyBraces
Since v3.0.0.
Default: spaces.inInterpolatedStringCurlyBraces = false
s"Hello ${ the } world!"
s"Hello ${ th.e } world!"
s"Hello ${ the() } world!"
s"Hello ${the} world!"
s"Hello ${ th.e} world!"
s"Hello ${the() } world!"
Config for this example:
spaces.inInterpolatedStringCurlyBraces = true
s"Hello ${oneHundred}% world!"
s"Hello ${th.e} world!"
s"Hello ${the()} world!"
s"Hello ${ oneHundred }% world!"
s"Hello ${ th.e} world!"
s"Hello ${the() } world!"
Config for this example:
spaces.inInterpolatedStringCurlyBraces = false
spaces.inParentheses
Default: spaces.inParentheses = false
foo( a, b )
foo(a, b)
Config for this example:
spaces.inParentheses=true
spaces.neverAroundInfixTypes
Default: spaces.neverAroundInfixTypes = []
def f: Foo##Repr
def g: Foo \/ Repr
// usage same operator not as type
def e = a ## b
def f: Foo##Repr
def g: Foo\/Repr
// usage same operator not as type
def e = a##b
Config for this example:
spaces.neverAroundInfixTypes=["##"]
spaces.aroundSymbolicInfixOperators
Added in v3.8.3, this parameter controls spaces around some infix operators
(unless spaces.neverAroundInfixTypes above
applies).
This parameter contains two subparameters, include and exclude, each either
a string containing a regex, or a list of regex alternatives. By default,
include is .* (i.e., matches everything) and exclude is ^$ (matches
nothing), thus all infix operators would use a space.
The logic is applied as follows:
- first, the infix operator must be symbolic (i.e., does not start with a letter or underscore) and not an assignment; otherwise, space is enforced;
- if the two characters on either side of the proposed space (i.e., the last character of the left-hand side and the first character of the operator, for spaces before the operator, or the last character of the operator and the first character of the right-hand side for spaces after the operator) can both be part of a symbolic operator, the space is enforced;
- otherwise, if the operator doesn't match any of the
includepatterns or matches one of theexcludepatterns, the space is not output.
def f: Foo##Repr
def g(a: Column, b: Column): Boolean = a===b || a ### b
def e(a: Int, b: Int) = a##b || a==b || a != b
def f: Foo ## Repr
def g(a: Column, b: Column): Boolean = a === b || a###b
def e(a: Int, b: Int) = a ## b || a == b || a!=b
Config for this example:
maxColumn = 80
spaces.aroundSymbolicInfixOperators.include = ".*" # default
spaces.aroundSymbolicInfixOperators.exclude = [ "^##$", "==" ]
spaces.afterKeywordBeforeParen
Default: spaces.afterKeywordBeforeParen = true
if(a) println("HELLO!")
while(a) println("HELLO!")
if (a) println("HELLO!")
while (a) println("HELLO!")
Config for this example:
spaces.afterKeywordBeforeParen = false
spaces.inByNameTypes
Default: spaces.inByNameTypes = true
def foo(a: =>A): A
def foo(a: => A): A
Config for this example:
spaces.inByNameTypes = false
spaces.afterSymbolicDefs
Default: spaces.afterSymbolicDefs = false
def +++ (a: A): F[A]
def +++(a: A): F[A]
Config for this example:
spaces.afterSymbolicDefs=true
spaces.beforeXxxArgInParens
Since v3.7.13.
These parameters control whether a space should be added before an argument of a function call or infix expression, if the argument is enclosed in parentheses.
They take the following values:
Never: no space is addedAlways: space is addedAfterSymbolic: space is added if the infix operator or function name is symbolic (doesn't start with a letter or underscore)
Please note that these parameters will not affect spacing after an
unary operator
(i.e., one of +, -, !, ~), as it's neither a function call nor an infix.
Also, spaces.beforeApplyArgInParens generalizes the special-case parameter
spaces.afterTripleEquals which only applies to a === function call.
# Defaults
spaces.beforeApplyArgInParens = "Never"
spaces.beforeInfixArgInParens = "Always"
+(baz)
=== (baz)
bar (baz)
foo + (baz)
foo === (baz)
foo bar (baz)
+(baz)
===(baz)
bar(baz)
foo +(baz)
foo ===(baz)
foo bar(baz)
Config for this example:
spaces.beforeApplyArgInParens = Always
spaces.beforeInfixArgInParens = Always
+(baz)
===(baz)
bar(baz)
foo +(baz)
foo ===(baz)
foo bar(baz)
+ (baz)
=== (baz)
bar (baz)
foo + (baz)
foo === (baz)
foo bar (baz)
Config for this example:
spaces.beforeApplyArgInParens = Never
spaces.beforeInfixArgInParens = Never
+(baz)
=== (baz)
bar(baz)
foo + (baz)
foo === (baz)
foo bar(baz)
+ (baz)
===(baz)
bar (baz)
foo +(baz)
foo ===(baz)
foo bar (baz)
Config for this example:
spaces.beforeApplyArgInParens = AfterSymbolic
spaces.beforeInfixArgInParens = AfterSymbolic
spaces.afterColonInMatchPattern
Since v3.8.3.
Default: spaces.afterColonInMatchPattern = "Always"
This parameter controls whether to output a space between a colon and a type in typed match patterns. It takes the following values:
Always:
foo match {
case e: Etype | f: Ftype =>
case g: Gtype =>
}
foo match {
case e: Etype | f: Ftype =>
case g: Gtype =>
}
Config for this example:
spaces.afterColonInMatchPattern = always
Never:
foo match {
case e:Etype | f:Ftype =>
case g:Gtype =>
}
foo match {
case e: Etype | f: Ftype =>
case g: Gtype =>
}
Config for this example:
spaces.afterColonInMatchPattern = never
NoAlternatives(when there are no pipe-separated alternatives):
foo match {
case e:Etype | f:Ftype =>
case g: Gtype =>
}
foo match {
case e: Etype | f: Ftype =>
case g: Gtype =>
}
Config for this example:
spaces.afterColonInMatchPattern = noAlternatives
Literals
Since v2.5.0.
Scalafmt allows flexible configuration of Integer and Floating Point literals formatting.
Default formatting:
123L
0xfff
0x1abL
10e-1
10e-1d
0b111
0b111L
123l
0XFff
0x1Abl
10E-1
10e-1D
0B111
0b111l
Each literals.* setting has three available options: Upper, Lower,
Unchanged.
literals.long
Default: literals.long = "Upper"
Responsible for the case of Long literals suffix L
123L
123l
Config for this example:
literals.long=Upper
literals.float
Default: literals.float = "Lower"
Responsible for the case of Float literals suffix F
42.0f
42.0F
Config for this example:
literals.float=Lower
literals.double
Default: literals.double = "Lower"
Responsible for the case of Double literals suffix D
42.0d
42.0d
Config for this example:
literals.double=Lower
literals.hexPrefix
Default: literals.hexPrefix = "Lower"
Responsible for the case of hex integer literals prefix 0x
0x123
0X123
Config for this example:
literals.hexPrefix=Lower
literals.hexDigits
Default: literals.hexDigits = "Lower"
Responsible for the case of hex integer literals digits
0xaaaa
0xaaaaL
0xaAaA
0xaAaAl
Config for this example:
literals.hexDigits=Lower
literals.long=Upper
literals.binPrefix
Since v3.8.4.
Default: literals.binPrefix = "Lower"
Responsible for the case of binary integer literals prefix 0b
0b111
0b111L
0B111
0B111l
Config for this example:
literals.binPrefix=Lower
literals.long=Upper
literals.scientific
Default: literals.scientific = "Lower"
Responsible for the case of Double literals exponent part
10E-1
10E-1f
10e-1
10e-1f
Config for this example:
literals.scientific=Upper
literals.float=Lower
XML
Controls formatting of Scala embedded within XML.
xmlLiterals.assumeFormatted
Since v2.6.0.
If set, formats embedded Scala relative to containing XML, making the assumption that XML itself is properly formatted. Otherwise, formatting is relative to the outer Scala code which contains the XML literals.
Default: xmlLiterals.assumeFormatted = false
object Example2 {
def apply() = {
<foo>
<bar>{
(1 + 2 + 3).toString(
"some long format"
)
}</bar>
</foo>
}
}
object Example2 {
def apply() = {
<foo>
<bar>{ (1 + 2 + 3).toString("some long format") }</bar>
</foo>
}
}
Config for this example:
maxColumn = 40
xmlLiterals.assumeFormatted = true
object Example2 {
def apply() = {
<foo>
<bar>{
(1 + 2 + 3).toString(
"some long format"
)
}</bar>
</foo>
}
}
object Example2 {
def apply() = {
<foo>
<bar>{ (1 + 2 + 3).toString("some long format") }</bar>
</foo>
}
}
Config for this example:
maxColumn = 40
xmlLiterals.assumeFormatted = false
Binpacking
binPack presets
One can specify the following values to binPack.preset:
neverorfalse:Alwaysortrue:Oneline:Onelinesjs:
The rest of parameters retain their default behaviour.
Literal argument lists
This group of parameters controls binpacking of an argument list if all arguments are considered to be literals. These parameters take precedence over forcing of config style.
The following parameters affect this behaviour:
binPack.literalArgumentLists: if false, this behaviour is disabled, other parameters ignoredbinPack.literalsMinArgCount: doesn't apply binpacking to calls with fewer argumentsbinPack.literals{Include,Exclude}: lists of regular expressions which define a literal- (since v2.5.0)
binPack.literalsIncludeSimpleExpr: allows a few selects (i.e.a.b), followed by a few nested single-argument apply calls, with literals as arguments- since v3.3.2, also includes
new
- since v3.3.2, also includes
- (since v2.5.0)
binPack.literalsSingleLine: the entire argument list will be formatted on one line, regardless ofmaxColumn
# Defaults
binPack.literalArgumentLists = true
binPack.literalsMinArgCount = 5
binPack.literalsInclude = [".*"]
binPack.literalsExclude = ["String", "Term.Name"]
binPack.literalsIncludeSimpleExpr = false
binPack.literalsSingleLine = false
val secret: List[Bit] = List(0, 0, 1, 1,
1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0,
1, 0, 0, 0, 1, 0, 1)
val secret: List[Bit] = List(0, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1)
Config for this example:
binPack.literalArgumentLists = true
val secret: List[Bit] = List(
0,
0,
1,
1,
1,
1,
1,
0,
0,
1,
1,
0,
1
)
val secret: List[Bit] = List(0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1)
Config for this example:
binPack.literalArgumentLists = false
val secret: List[Bit] = List(
0, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1)
val secret: List[Bit] = List(0, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1,
0, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1)
Config for this example:
binPack.literalArgumentLists = true
binPack.literalsSingleLine = true
binPack.parentConstructors
Parent constructors are B (since 3.4.1), C and D in class A extends B with C and D. Changed
from a boolean to a wider set of options in v2.6.0.
Default: binPack.parentConstructors = "source"
The behaviour of binPack.parentConstructors = source depends on the value of
newlines.source; keep maps to keep and attempts to preserve the
space if there's no line break in the source, fold maps to Oneline, rest to Never.
binPack.parentConstructors=Always
This option attempts to binpack parents, formatting as many on each line as will fit.
object A {
trait Foo
extends Bar with Baz
trait Foo
extends Bar with Baz
}
object A {
trait Foo
extends Bar
with Baz
trait Foo extends Bar with Baz
}
Config for this example:
binPack.parentConstructors = Always
maxColumn = 30
binPack.parentConstructors=Never
This option will attempt to format the entire declaration on one line (starting
with class or trait and including all parents); otherwise, will enforce
breaks before each.
object A {
trait Foo
extends Bar
with Baz
}
object A {
trait Foo extends Bar with Baz
}
Config for this example:
binPack.parentConstructors = Never
maxColumn = 30
binPack.parentConstructors=Oneline
This option will attempt to format all parents on one line (starting with
extends and including all parents); otherwise, will enforce breaks before each.
object A {
class Foo(a: Int)
extends Bar with Baz
class Foo(
a: Int
) extends Bar with Baz
}
object A {
class Foo(a: Int)
extends Bar
with Baz
class Foo(
a: Int
)
extends Bar
with Baz
}
Config for this example:
binPack.parentConstructors = Oneline
maxColumn = 30
binPack.parentConstructors=OnelineIfPrimaryOneline
This option will attempt to format all parents on one line (from extends and
including all parents), but only if the primary constructor fits on one line
as well (the same or previous); otherwise, will enforce breaks before each.
object A {
class Foo(a: Int, b: Int)
extends Bar with Baz
class Foo(
a: Int,
b: Int
) extends Bar
with Baz
}
object A {
class Foo(a: Int, b: Int)
extends Bar
with Baz
class Foo(
a: Int,
b: Int
) extends Bar with Baz
}
Config for this example:
binPack.parentConstructors = OnelineIfPrimaryOneline
maxColumn = 30
binPack.parentConstructors=keep
This option attempts to preserve breaks before each parent.
object A {
class Foo(a: Int, b: Int) extends Bar(
a,
b
) with Baz
with Qux
class Foo(a: Int, b: Int)
extends Bar(
a,
b
) with Baz
with Qux
}
object A {
class Foo(a: Int, b: Int) extends Bar(
a,
b
) with Baz
with Qux
class Foo(a: Int, b: Int)
extends Bar(
a,
b
) with Baz
with Qux
}
Config for this example:
binPack.parentConstructors = keep
binPack.parentConstructors=ForceBreak
This option will enforce a break before each parent. As usual, the break is only actually introduced if indented position on the next line is less than the current. Added in 3.8.4.
object A {
class Foo(a: Int)
extends Bar
with Baz
class Foo(
a: Int,
b: Int,
c: String,
d: Double
) extends Bar
with Baz
}
object A {
class Foo(a: Int) extends Bar with Baz
class Foo(a: Int, b: Int, c: String, d: Double) extends Bar with Baz
}
Config for this example:
binPack.parentConstructors = ForceBreak
maxColumn = 45
binPack.xxxSite
Controls binpacking around method/type definition sites (binPack.defnSite) or
method call sites (binPack.callSite) (both were called
unsafeXxxSite up until v3.8.1). The following parameter values are supported
since v3.0.0:
Neverdisables the functionality (also takesfalse)Alwaysenables the functionality (also takestrue)Onelineensures multiline arguments are not binpacked, that is, they cannot be on the same line as any other argument; also, if the multiline argument is last and the call site is followed by either a curried argument clause or a chained method, it cannot be on the same line as the next identifier after the closing delimiterOnelineSjsensures multiline arguments are not binpacked, likeOnelineabove, but ignores line breaks within expressions in braces
Additionally, when binpacking is enabled (i.e., not Never), some nested
expressions in brackets (if the binpacked clause uses brackets) or in braces
(if it uses parentheses) wouldn't be indented as would happen in usual cases;
this logic is always used for OnelineSjs, and only for single-element clauses
otherwise.
When not disabled, these parameters have complex interactions with
newline.source,
newlines.configStyle.xxxSite.prefer
(aka cfgStyle below) and
danglingParentheses.xxxSite (aka dangle).
Keep in mind that when config-style is forced,
it takes precedence over options described below.
The code that is being binpacked might also yield an additional force over
the selection of breaks, such as when dealing with multiline arguments
(they behave differently under Oneline setting, or depending on
alignment).
newlines.source=classiccfgStyle=T,dangle=T: use cfg-style if both parens had breaks, otherwise binpack without breaks- before v3.8.2, this formatting was used for
callSitewithdangle=Fas well
- before v3.8.2, this formatting was used for
cfgStyle=T,dangle=F: (scala.js) use cfg-style if close paren had a break; otherwise, binpack without breakscfgStyle=F,dangle=T: binpack; if both parens had breaks, keep; otherwise, use no breaks- before v3.8.2, this formatting was used for
defnSitewithcfgStyle=Tand anydangle
- before v3.8.2, this formatting was used for
cfgStyle=F,dangle=F: binpack without breaks- before v3.8.2, this formatting was used for
callSitewithcfgStyle=Fand anydangle, and fordefnSitewithcfgStyle=Fanddangle=F
- before v3.8.2, this formatting was used for
newlines.source=keep: always preserve break after open parencfgStyle=T,dangle=T: use cfg-style if open paren had a break; otherwise, binpack and keep close breakcfgStyle=T,dangle=F: (scala.js) use cfg-style if close paren had a break; otherwise, binpack and keep close breakcfgStyle=F,dangle=T: binpack; force close break if open paren had a break; otherwise, preserve itcfgStyle=F,dangle=F: binpack; preserve close break
newlines.source=fold: if single line is not possible:cfgStyle=T,dangle=T: binpack with both breakscfgStyle=T,dangle=F: binpack with dangling open and tucked closecfgStyle=F,dangle=T: binpack with tucked open and dangling close- if
binPack.indentCallSiteOnceis set, we will not force dangling as it might lead to consecutive lines with a closing parenthesis at the same indentation level
- if
cfgStyle=F,dangle=F: binpack without breaks
newlines.source=unfold: if single line is not possible:- open dangles, close break matches
dangle,cfgStyleis ignored
- open dangles, close break matches
Please also see callSite indentation parameters.
binPack.bracketXxxSite
Since v3.0.4.
If set explicitly, will be used for type arguments or parameters,
instead of the respective binPack.xxxSite.
binPack.importSelectors
Import selectors (those grouped in {...}) will always be formatted on a single
line if they fit without exceeding maxColumn. This parameter controls how they
will be handled if they overflow.
(Prior to v3.8.4, it was called importSelectors.)
Default: binPack.importSelectors = "unfold"
Takes the following parameters:
unfold: format one per line (prior to v3.8.4, callednoBinPack)fold: fit as many as possible on each line (prior to v3.8.4, calledbinPack)singleLine: format all on one line
import a.b.{
c,
d,
e,
f,
g
}
import a.b.{c, d, e, f, g}
Config for this example:
maxColumn = 10
binPack.importSelectors = unfold
import a.b.{
c, d, e,
f, g
}
import a.b.{c, d, e, f, g}
Config for this example:
maxColumn = 10
binPack.importSelectors = fold
import a.b.{c, d, e, f, g}
import a.b.{c, d, e, f, g}
Config for this example:
maxColumn = 10
binPack.importSelectors = singleLine
Select chains
The parameters below control formatting of select chains, such as which select expressions are considered to start a chain.
Generally, a chain can either be formatted on one line up to the last select, or will have a break on the first select.
when
newlines.selectChains.style
This parameter controls how select chains (sequences of .method invocations)
are formatted.
It takes the same values as newlines.source; use null
(default) to fall back on the current value of newlines.source.
Since v3.0.0.
keep: attempts to preserve breakfold: attempts to avoid breaksunfold: forces breaks on each select unless all fit on a single lineclassic(i.e.,nullandnewlines.sourceis not specified).
newlines.selectChains.enclose
- Added in v2.6.2
- Before v3.8.4, was called
optIn.encloseClassicChainsand applied only tostyle = classic- By default, enabled unless
style = classic
Controls what happens if a chain enclosed in parentheses is followed by additional selects. Those additional selects will be considered part of the enclosed chain if and only if this flag is false.
Default: newlines.selectChains.enclose = null
(foo.map(_ + 1).map(_ + 1))
.filter(_ > 2)
(foo.map(_ + 1).map(_ + 1))
.filter(_ > 2)
Config for this example:
newlines.selectChains.enclose = true
maxColumn = 30
(foo
.map(_ + 1)
.map(_ + 1))
.filter(_ > 2)
(foo.map(_ + 1).map(_ + 1))
.filter(_ > 2)
Config for this example:
newlines.selectChains.enclose = false
maxColumn = 30
newlines.selectChains.classicCanStartWithBraceApply
- Prior to v3.8.4, was called
includeCurlyBraceInSelectChains- Applies only if
style = classic(always enabled for others)
Controls if select followed by curly braces can start a chain.
Default: newlines.selectChains.classicCanStartWithBraceApply = true
List(1)
.map { x =>
x + 2
}
.filter(_ > 2)
List(1).map { x =>
x + 2
}
.filter(_ > 2)
Config for this example:
newlines.selectChains.classicCanStartWithBraceApply = true
List(1).map { x =>
x + 2
}
.filter(_ > 2)
List(1)
.map { x =>
x + 2
}
.filter(_ > 2)
Config for this example:
newlines.selectChains.classicCanStartWithBraceApply = false
newlines.selectChains.classicCanStartWithoutApply
- Prior to v3.8.4, was called
includeNoParensInSelectChains- Applies only if
style = classic(always enabled for others)
Controls if select not followed by an apply can start a chain.
Default: newlines.selectChains.classicCanStartWithoutApply = false
List(1)
.toIterator
.buffered
.map(_ + 2)
.filter(_ > 2)
List(1).toIterator.buffered
.map(_ + 2)
.filter(_ > 2)
Config for this example:
newlines.selectChains.classicCanStartWithoutApply = true
List(1).toIterator.buffered
.map(_ + 2)
.filter(_ > 2)
List(1).toIterator.buffered.map(_ + 2).filter(_ > 2)
Config for this example:
newlines.selectChains.classicCanStartWithoutApply = false
newlines.selectChains.classicKeepFirst
- Prior to v3.8.4, was called
optIn.breakChainOnFirstMethodDot- Applies only if
style = classic
Keeps the break on the first select of the chain if the source contained one. Has no effect if there was no newline in the source.
Default: newlines.selectChains.classicKeepFirst = true
// collapse into a single line
foo.map(_ + 1).filter(_ > 2)
// collapse into a single line
foo
.map(_ + 1)
.filter(_ > 2)
Config for this example:
newlines.selectChains.classicKeepFirst = false
// preserve break on first dot and break on subsequent dots
foo
.map(_ + 1)
.filter(_ > 2)
// preserve break on first dot and break on subsequent dots
foo
.map(_ + 1).filter(_ > 2)
Config for this example:
newlines.selectChains.classicKeepFirst = true
newlines.selectChains.classicKeepAfterFirstBreak
- Prior to v3.8.4, was called
optIn.breaksInsideChains- Applies only if
style = classic
Controls whether to preserve a newline before each subsequent select when the very first one used a line break; that is, this parameter doesn't prohibit single-line formatting even if there are source breaks down the chain.
If false, each subsequent select within the chain will behave exactly like the first, that is, either the entire chain will be formatted on one line, or will contain a break on every select.
If true, preserves existence or lack of breaks on subsequent selects if the first select was formatted with a newline.
Default: newlines.selectChains.classicKeepAfterFirstBreak = false
foo
.bar(_ + 1)
.baz(_ > 2).qux
foo.bar(_ + 1).baz(_ > 2).qux
foo
.bar(_ + 1).baz(_ > 2).qux(
_ * 12
)
foo.bar(_ + 1).baz(_ > 2).qux {
_ * 12
}
foo
.bar(_ + 1)
.baz(_ > 2).qux(_ * 12)
foo.bar(_ + 1)
.baz(_ > 2).qux
foo.bar(_ + 1).baz(_ > 2).qux
foo.bar(_ + 1).baz(_ > 2).qux(_ * 12)
foo.bar(_ + 1).baz(_ > 2).qux { _ * 12 }
foo.bar(_ + 1)
.baz(_ > 2).qux(_ * 12)
Config for this example:
newlines.selectChains.classicKeepAfterFirstBreak = true
maxColumn = 35
foo
.bar(_ + 1)
.baz(_ > 2)
.qux
foo.bar(_ + 1).baz(_ > 2).qux
foo
.bar(_ + 1)
.baz(_ > 2)
.qux(_ * 12)
foo.bar(_ + 1).baz(_ > 2).qux {
_ * 12
}
foo
.bar(_ + 1)
.baz(_ > 2)
.qux(_ * 12)
foo.bar(_ + 1)
.baz(_ > 2).qux
foo.bar(_ + 1).baz(_ > 2).qux
foo.bar(_ + 1).baz(_ > 2).qux(_ * 12)
foo.bar(_ + 1).baz(_ > 2).qux { _ * 12 }
foo.bar(_ + 1)
.baz(_ > 2).qux(_ * 12)
Config for this example:
newlines.selectChains.classicKeepAfterFirstBreak = false
maxColumn = 35
Miscellaneous
lineEndings
This parameter controls which end-of-line character sequences will be output, takes the following values:
unix: outputs LF (U+000A)windows: uses CRLF (U+000D U+000A)preserve: if an input file contains CRLF anywhere, use CRLF for every line; otherwise, LF
By default, this parameter is assumed to be set to unix. However, since
v3.8.3, if the formatter is invoked on Windows, project.git
is set and git parameter
core.autocrlf
is configured, then default value will be changed to windows if
autocrlf=true, and preserve if false.
Markdown Formatting
Since v3.0.0.
Will format all scala mdoc fences inside Markdown files.
# Format default filetypes + Markdown files
project.includePaths."+" = ["glob:**.md"]
# *Only* format Markdown files
project.includePaths = [ "glob:**.md" ]
Before:
Markdown prose beginning.
```scala mdoc
val x = 3
```
Markdown prose end.
After:
Markdown prose beginning.
```scala mdoc
val x = 3
```
Markdown prose end.
Advanced: formatting process
Currently, the formatting process consists of several stages:
- code parsing using scalameta parser
- keep in mind that any parsing errors output by scalafmt are in fact coming from scalameta
- format-agnostic rewrites
- during this step, we apply most of the rewrites rules (except those which depend on how we decided to format the code)
- preparing whitespace splits
- we iterate over all pairs of adjacent non-whitespace tokens and identify possible options for the output whitespace between them
- selection is between
- a newline with a possible subsequent indentation
- a single space
- no space at all
- each option also comes with a penalty and a possible policy on how to handle
subsequent splits
- for instance, this is how we ensure single-line formatting, by adding a policy which removes all newline splits before a given end token
- route search
- we look for an optimal path through the space of non-whitespace tokens and splits between them
- each partial state includes, among other things:
- the last token considered
- the split selection preceding it
- total cumulative penalty
- the partial state for the previous token
- we maintain a priority queue containing these partial states, by preferring states with lower penalty and then higher token offset (i.e. those which are closer to the completion)
- at each iteration, we select a state from that priority queue (we'll call this event a state visit), apply all available non-expired policies to available subsequent whitespace splits, insert the next state for each of them back into the priority queue
- we stop as soon as the top state in the priority queue is completed (i.e.,
refers to the
EOFtoken)
- output
- after we have obtained our final formatting state, we need to write out the code applying the splits selected during the route search
- as part of this final stage, we perform a few steps:
- apply format-dependent rewrites (as these frequently need to know how
many line breaks we have selected to output for a given code passage);
for instance,
- inserting braces,
- adding or removing end markers
- modifying blank lines
- rewrite comments and docstrings if configured
- rewrite trailing commas: as part of this rule, we remove any trailing commas during the format-agnostic rewrites stage and add them as needed during this stage, if our splits solution ended up with a line break before the closing paren or brace
- alignment: determine any additional horizontal space to be added
- if alignment is for some reason is not possible, we don't go back to the previous stage (route search) to determine a different splits solution which might possibly make alignment possible
- this might happen, for instance, if alignment leads to disallowed line overflow
- apply format-dependent rewrites (as these frequently need to know how
many line breaks we have selected to output for a given code passage);
for instance,
Advanced: Route search optimizations
There are several configuration parameters (most of them under the
runner.optimizer section) which control some aspects of the algorithm during
this formatting stage.
Route search optimizations: giving up
The following parameters control when we abort the search and throw a
SearchStateExploded exception:
runner.maxStateVisits: when the total number of state visits exceeds this valuerunner.optimizer.maxVisitsPerToken: when the number of state visits for a given token exceeds this value (also see below)runner.optimizer.escapeInPathologicalCases: if this flag is disabled, ignoremaxVisitsPerTokenabove
# Defaults
runner.maxStateVisits = null
runner.optimizer.maxVisitsPerToken = 10000
runner.optimizer.escapeInPathologicalCases = true
Route search optimizations: discarding some states
Normally, the search algorithm considers any state as a possible intermediate state which might complete a valid solution.
However, in some cases we might decide to require certain conditions to be met for a state to be considered. A state need to qualify under at least one of the checks controlled by these parameters:
runner.optimizer.pruneSlowStates:- accepts
trueorfalse; also, since v3.8.4,No,YesorOnly - if this flag is disabled (
falseorNo), any state qualifies - if it is enabled:
- if the search algorithm is ultimately unable to find a completed solution
(because some states might be discarded) and if the value is not
Only, then it will disable the flag and make another attempt - during the first run, when the flag is enabled, a state qualifies if it is not "worse" than a previously recorded "best" state for this token
- the "best" state is the first state with a newline split at the given token (recursive calls over subsequences are excluded, see sections below)
- a state is "worse" when it has higher penalty and wider indentation
- if the search algorithm is ultimately unable to find a completed solution
(because some states might be discarded) and if the value is not
- accepts
runner.optimizer.disableOptimizationsInsideSensitiveAreas: a state qualifies if this flag is disabled or the current token is within a so-called no-optimization zone (and if the state satisfies these criteria, we'll call it a no-optimization state):- between matching parentheses of method call (either those enclosing the method arguments, or the entire call itself)
- between matching braces of a type refinement
# Defaults
runner.optimizer.disableOptimizationsInsideSensitiveAreas = true
runner.optimizer.pruneSlowStates = "Yes"
Route search optimizations: preferential treatment
In some cases, we modify the search logic to boost priority of some states. The following sections describe when and how this might be done.
Route search optimizations: optimal hints
Some splits might be marked as starting a token subsequence with an optimal end token (for example, a segment which must be formatted without line breaks).
In such cases, if flag runner.optimizer.acceptOptimalAtHints is enabled and
the hint is deemed valid (see below), the route search algorithm calls itself
recursively with the state that ends in this "optimal" split and just on this
shorter subsequence, returning the first state which completed this subsequence,
with the following restrictions:
- the state penalty at the end must be the same as at the beginning, so no splits within the subsequence can have non-zero penalty
- the so-called nested depth is incremented (set to zero initially) and then
must not exceed
runner.optimizer.maxDepth
There are several checks that are performed to determine validity of the hint:
- no optimal subsequence has yet been identified for the current state
- there are multiple possible next states; we don't do this if there's only one
- the split has zero penalty, including from any unexpired policies
# Defaults
runner.optimizer.acceptOptimalAtHints = true
runner.optimizer.maxDepth = 100
Route search optimizations: longer blocks
Another form of getting ahead faster is by recursing on a long-enough block
(whether enclosed in explicit braces or by using significant indentation),
where "long enough" is loosely defined as having sufficient non-whitespace
characters to fill at least three dense maxColumn lines.
Unlike the case with optimal hints above, there's no restriction on the penalty added by the splits within the block.
This optimization is enabled when:
runner.optimizer.recurseOnBlocksis enabled, and- the token is within a no-optimization zone.
Default: runner.optimizer.recurseOnBlocks = true
Route search optimizations: statements
Another way to boost priority is simply by adding a new priority queue, so that any earlier states, from the existing priority queue, wouldn't be considered until the new one is exhausted.
We might create a new priority queue if we encounter a new block statement, and this optimization is enabled when:
runner.optimizer.dequeueOnNewStatementsis enabled, and- we are in a recursive call on a subsequence or the token is not within a no-optimization zone.
Default: runner.optimizer.dequeueOnNewStatements = true
Route search optimizations: arg or param clause
Similar to statements above, we might create a new priority queue when we encounter a long-enough call-site argument or defn-site parameter clause.
# Defaults
runner.optimizer.callSite = {"minSpan": 150, "minCount": 2}
runner.optimizer.defnSite = {"minSpan": -1, "minCount": 0}
This optimization is enabled when all these criteria are satisfied:
xxxSite.minSpan: must be non-negative and not exceed the character span covered by the entire clause, excluding any whitespace- prior to v3.8.1, this parameter was called
forceConfigStyleOnOffset - prior to v3.8.4, the parameter had to be strictly less than the span, and the span calculation excluded the last token of the clause (a closing delim spanning a single-character, unless optional braces had been used)
- prior to v3.8.1, this parameter was called
xxxSite.minCount: must be positive and not exceed the number of arguments (prior to v3.8.2, this parameter was calledforceConfigStyleMinCount)
These parameters cannot be overridden within a file
Edition
Removed in 2.7.0
Editions are no longer used. They were kept for backwards compatibility with old
configuration files but new changes to the default Scalafmt formatting behavior
will not respect the edition setting.
Other
To find all available configuration options, it's best to browse the source code
of Scalafmt. A good place to start is ScalafmtConfig. Observe that this
listing below is the top-level, there are more configuration options if you
visited nested fields like spaces and newlines.
The values for parameters below are purely for informational purposes. They use pseudo-formatting similar to but incompatible with
scalafmt.conf. Some of them show values which can't be explicitly specified.
version = "3.8.4-RC4"
maxColumn = 80
docstrings.oneline = keep
docstrings.removeEmpty = false
docstrings.wrap = unfold
docstrings.wrapMaxColumn = null
docstrings.forceBlankLineBefore = true
docstrings.blankFirstLine = null
docstrings.style = SpaceAsterisk
comments.wrap = no
comments.wrapSingleLineMlcAsSlc = false
comments.wrapStandaloneSlcAsSlc = false
binPack.callSite = Never
binPack.defnSite = Never
binPack.bracketCallSite = null
binPack.bracketDefnSite = null
binPack.indentCallSiteOnce = false
binPack.indentCallSiteSingleArg = true
binPack.parentConstructors = source
binPack.literalArgumentLists = true
binPack.literalsIncludeSimpleExpr = false
binPack.literalsSingleLine = false
binPack.literalsMinArgCount = 5
binPack.literalsInclude = [
".*"
]
binPack.literalsExclude = [
String
"Term.Name"
]
binPack.importSelectors = unfold
indent.main = 2
indent.significant = null
indent.callSite = 2
indent.ctrlSite = null
indent.binPackCallSite = null
indent.defnSite = 4
indent.binPackDefnSite = null
indent.caseSite = 4
indent.matchSite = null
indent.ctorSite = null
indent.extraBeforeOpenParenDefnSite = 0
indent.relativeToLhsLastLine = []
indent.fewerBraces = never
indent.afterInfixSite = null
indent.extendSite = 4
indent.withSiteRelativeToExtends = 0
indent.commaSiteRelativeToExtends = 2
indent.yieldKeyword = true
indent.infix.exemptScope = oldTopLevel
indent.infix.includeRegex = ".*"
indent.infix.excludeRegex = "^(&&|\|\|)$"
align.allowOverflow = false
align.delayUntilSpace = true
align.multiline = false
align.stripMargin = true
align.closeParenSite = false
align.openBracketCallSite = null
align.openParenCallSite = false
align.openParenCtrlSite = false
align.openBracketDefnSite = null
align.openParenDefnSite = false
align.openParenTupleSite = null
align.beforeOpenParenDefnSite = false
align.beforeOpenParenCallSite = false
align.inInterpolation = false
align.tokens = [
{
code = "=>"
owners = [
{
regex = Case
parents = []
}
]
}
]
align.arrowEnumeratorGenerator = false
align.treeCategory."Defn.Trait" = "class/object/trait/enum"
align.treeCategory."Defn.Object" = "class/object/trait/enum"
align.treeCategory."Defn.Val" = "given/val/var/def"
align.treeCategory."Defn.Enum" = "class/object/trait/enum"
align.treeCategory."Defn.Macro" = "given/val/var/def"
align.treeCategory."Decl.Def" = "given/val/var/def"
align.treeCategory."Defn.Def" = "given/val/var/def"
align.treeCategory."Defn.GivenAlias" = "given/val/var/def"
align.treeCategory."Defn.Var" = "given/val/var/def"
align.treeCategory."Enumerator.Generator" = for
align.treeCategory."Enumerator.Val" = for
align.treeCategory."Defn.Class" = "class/object/trait/enum"
spaces.beforeContextBoundColon = Never
spaces.beforeApplyArgInParens = Never
spaces.beforeInfixArgInParens = Always
spaces.afterTripleEquals = false
spaces.inImportCurlyBraces = false
spaces.inInterpolatedStringCurlyBraces = false
spaces.inParentheses = false
spaces.neverAroundInfixTypes = []
spaces.aroundSymbolicInfixOperators = null
spaces.afterKeywordBeforeParen = true
spaces.inByNameTypes = true
spaces.afterSymbolicDefs = false
spaces.afterColonInMatchPattern = Always
literals.long = Upper
literals.float = Lower
literals.double = Lower
literals.hexDigits = Lower
literals.hexPrefix = Lower
literals.binPrefix = Lower
literals.scientific = Lower
lineEndings = null
rewrite.rules = []
rewrite.scala3.convertToNewSyntax = false
rewrite.scala3.newSyntax.control = true
rewrite.scala3.newSyntax.deprecated = true
rewrite.scala3.removeOptionalBraces.enabled = false
rewrite.scala3.removeOptionalBraces.fewerBracesMinSpan = 2
rewrite.scala3.removeOptionalBraces.fewerBracesMaxSpan = 0
rewrite.scala3.removeOptionalBraces.fewerBracesParensToo = false
rewrite.scala3.removeOptionalBraces.oldSyntaxToo = false
rewrite.scala3.countEndMarkerLines = all
rewrite.scala3.removeEndMarkerMaxLines = 0
rewrite.scala3.insertEndMarkerMinLines = 0
rewrite.insertBraces.minLines = 0
rewrite.insertBraces.allBlocks = false
rewrite.redundantBraces.defnBodies = all
rewrite.redundantBraces.includeUnitMethods = true
rewrite.redundantBraces.maxBreaks = 100
rewrite.redundantBraces.stringInterpolation = false
rewrite.redundantBraces.oneStatApply.parensMaxSpan = 0
rewrite.redundantBraces.oneStatApply.bracesMinSpan = -1
rewrite.redundantBraces.generalExpressions = true
rewrite.redundantBraces.ifElseExpressions = false
rewrite.redundantParens.infixSide = null
rewrite.sortModifiers.order = [
implicit
final
sealed
abstract
override
private
protected
erased
lazy
open
transparent
inline
infix
opaque
]
rewrite.imports.sort = none
rewrite.imports.expand = false
rewrite.imports.contiguousGroups = only
rewrite.imports.groups = []
rewrite.preferCurlyFors.removeTrailingSemicolonsOnly = false
rewrite.trailingCommas.allowFolding = true
rewrite.trailingCommas.style = never
rewrite.avoidInfix.includeFilters = [
"[\w\d_]+"
]
rewrite.avoidInfix.excludeFilters = [
until
to
by
eq
ne
]
rewrite.avoidInfix.excludeScalaTest = null
rewrite.avoidInfix.excludePlaceholderArg = true
rewrite.avoidInfix.excludePostfix = false
rewrite.avoidInfix.excludeMatch = true
newlines.source = null
newlines.avoidInResultType = false
newlines.beforeTypeBounds = null
newlines.neverBeforeJsNative = false
newlines.sometimesBeforeColonInMethodReturnType = true
newlines.beforeOpenParenDefnSite = null
newlines.beforeOpenParenCallSite = null
newlines.penalizeSingleSelectMultiArgList = true
newlines.beforeCurlyLambdaParams = never
newlines.topLevelStatementBlankLines = []
newlines.topLevelStatementsMinBreaks = 1
newlines.topLevelStatements = []
newlines.beforeTemplateBodyIfBreakInParentCtors = false
newlines.topLevelBodyIfMinStatements = []
newlines.topLevelBodyMinStatements = 2
newlines.afterCurlyLambdaParams = never
newlines.implicitParamListModifierForce = []
newlines.implicitParamListModifierPrefer = null
newlines.alwaysBeforeElseAfterCurlyIf = false
newlines.forceBeforeAssign = never
newlines.forceBeforeMultilineAssign = null
newlines.beforeMultiline = null
newlines.beforeMultilineDef = null
newlines.selectChains.style = null
newlines.selectChains.enclose = null
newlines.selectChains.classicKeepFirst = true
newlines.selectChains.classicKeepAfterFirstBreak = false
newlines.selectChains.classicCanStartWithBraceApply = true
newlines.selectChains.classicCanStartWithoutApply = false
newlines.afterInfix = null
newlines.afterInfixBreakOnNested = false
newlines.afterInfixMaxCountPerExprForSome = 10
newlines.afterInfixMaxCountPerFile = 500
newlines.avoidForSimpleOverflow = []
newlines.inInterpolation = allow
newlines.ignoreInSyntax = true
newlines.avoidAfterYield = true
newlines.configStyle.callSite = null
newlines.configStyle.defnSite = null
newlines.configStyle.bracketCallSite = null
newlines.configStyle.bracketDefnSite = null
newlines.configStyle.fallBack.prefer = true
newlines.configStyle.fallBack.forceIfOptimized = null
newlines.configStyle.beforeComma = false
newlines.annotation = true
newlines.selfAnnotation = true
runner.debug = false
runner.completeCallback = "<FormatEvent => Unit>"
runner.eventCallback = "<FormatEvent => Unit>"
runner.parser = Source
runner.optimizer.dequeueOnNewStatements = true
runner.optimizer.escapeInPathologicalCases = true
runner.optimizer.maxVisitsPerToken = 10000
runner.optimizer.maxDepth = 100
runner.optimizer.acceptOptimalAtHints = true
runner.optimizer.disableOptimizationsInsideSensitiveAreas = true
runner.optimizer.pruneSlowStates = Yes
runner.optimizer.recurseOnBlocks = true
runner.optimizer.callSite.minSpan = 150
runner.optimizer.callSite.minCount = 2
runner.optimizer.defnSite.minSpan = -1
runner.optimizer.defnSite.minCount = 0
runner.maxStateVisits = null
runner.dialect = null
runner.ignoreWarnings = false
runner.fatalWarnings = false
assumeStandardLibraryStripMargin = false
danglingParentheses.callSite = true
danglingParentheses.defnSite = true
danglingParentheses.ctrlSite = true
danglingParentheses.tupleSite = null
danglingParentheses.bracketCallSite = null
danglingParentheses.bracketDefnSite = null
danglingParentheses.exclude = null
verticalMultiline.atDefnSite = false
verticalMultiline.arityThreshold = 100
verticalMultiline.newlineAfterOpenParen = false
onTestFailure = ""
encoding = "UTF-8"
project.git = false
project.layout = null
project.includePaths = [
"glob:**.scala"
"glob:**.sbt"
"glob:**.sc"
"glob:**.mill"
]
project.excludePaths = []
project.includeFilters = []
project.excludeFilters = []
xmlLiterals.assumeFormatted = false
formatOn = [
"@formatter:on"
"format: on"
]
formatOff = [
"@formatter:off"
"format: off"
]