Changelog
Ce contenu n’est pas encore disponible dans votre langue.
@biomejs/biome
Section titled “@biomejs/biome”Major Changes
Section titled “Major Changes”-
Biome now resolves globs and paths from the configuration. Before, paths and globs were resolved from the working directory.
-
Biome now raises a warning diagnostic for suppression comments that have
<explanation>
as reason.<explanation>
is provided as a placeholder when applying the suppression code fix from LSP editors. -
Removed the
--config-path
argument from thebiome lsp-proxy
andbiome start
commands.The option was overriding the configuration path for all workspaces opened in the Biome daemon, which led to a configuration mismatch problem when multiple projects are opened in some editors or IDEs.
If you are using one of our official plugins for IDEs or editors, it is recommended to update it to the latest version of the plugin, or you will get unexpected behavior.
If you are a developer of a plugin, please update your plugin to use the
workspace/configuration
response instead of using the--config-path
argument. Biome’s LSP will resolve a configuration in the workspace automatically, so it is recommended to keep it empty unless you are using a custom configuration path. -
Downgraded some code fixes to unsafe which were previously safe.
The following rules have now a unsafe fix:
If you want to keep applying these fixes automatically, configure the rule fix as safe:
{"linter": {"rules": {"correctness": {"noFlatMapIdentity": {"level": "error","fix": "safe"},"noUnusedImports": {"level": "error","fix": "safe"}}}}} -
Previously the lint rules
noControlCharactersInRegex
andnoMisleadingCharacterClass
checked both regular expression literals like/regex/
and dynamically built regular expressions likenew RegExp("regex")
.Checking dynamically built regular expressions has many limitations, edge cases, and complexities. In addition, other rules that lint regular expressions don’t check dynamically built regular expressions.
Rather than add support for other rules and have half-baked checking, we decided to remove support for dynamically built regular expressions.
Now the lint rules
noControlCharactersInRegex
andnoMisleadingCharacterClass
only check literals of regular expressions. -
The lint rule
noRestrictedGlobals
now supports customizing message for each global name.For example, the following configuration:
{"options": {"deniedGlobals": {"$": "jQuery is not allowed. Use native DOM manipulation instead."}}}emits a diagnostic:
index.js:1:13 lint/style/noRestrictedGlobals ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━⚠ Do not use the global variable $.> 1 │ console.log($)│ ^2 │ℹ jQuery is not allowed. Use native DOM manipulation instead.Breaking Change: The option
deniedGlobals
is now a record instead of an array. Runbiome migrate
to migrate the configuration automatically. -
Removed
include
andignore
fields in favor of the new fieldincludes
.The Biome configuration file allows users to specify which files should be processed using glob patterns. Prior to Biome 2.0, this was done using the
include
andignore
fields. In Biome 2.0,include
andignore
are removed and replaced byincludes
. You can runbiome migrate
to convertinclude
andignore
intoincludes
automatically.includes
uses a different glob pattern format that fixes many issues and many other limitations that Biome users reported.includes
accepts an array of glob patterns. A glob pattern starting with a!
is a negated pattern also called exception. This replacesignore
patterns and allows users to create chains of include and ignore patterns. Thus, it is now possible to include again a file previously ignored. This was not possible withinclude
andignore
, becauseignore
has priority overinclude
.The semantics of
*
and**/*
have changed too. Before, withinclude
andignore
, the glob*
was interpreted as**/*
. Now, withincludes
, the globs*
and**/*
are interpreted differently. The first pattern matches all files that are inside a folder. The second pattern recursively matches all files and sub-folders inside a folder.Let’s take an example. Given the following file hierarchy of a project…
├── biome.json├── src│ ├── file.js│ ├── file.ts│ ├── out.gen.js│ ├── file.test.js│ └── test│ └── special.test.js└── test ...…we want:
- Ignore all files ending with
.test.js
, exceptspecial.test.ts
. - Ignore all files of the
test
directory. Thetest
directory is located at the root of the project. - Execute the linter on files in the
src
directory, that don’t end with.gen.js
. Thesrc
directory is located at the root of the project. - Enable the
noDefaultExport
lint rule on files ending with.ts
.
Prior to Biome 2.0, the configuration might look like:
{"files": {"ignore": ["*.test.js", "test"]},"linter": {"include": ["src/**"],"ignore": ["*.gen.js"],"enabled": true},"overrides": [{"include": ["*.ts"],"linter": { "rules": { "style": { "noDefaultExport": "on" } } }}]}Unfortunately, the configuration doesn’t quite fit what we want:
- There is no way to ignore files and unignore one of them.
Thus, we ignore all files ending with
.test.js
, includingspecial.test.ts
. - The configuration ignores all directories named
test
, includingsrc/test
. - The linter is executed on all files of all directories named
src
All these issues and limitations are fixed with
includes
. Here the migrated configuration:{"files": {"includes": ["**", "!**/*.test.js", "**/special.test.ts", "!test"]},"linter": {"includes": ["src/**", "!**/*.gen.js"],"enabled": true},"overrides": [{"includes": ["**/*.ts"],"linter": { "rules": { "style": { "noDefaultExport": "on" } } }}]}- All files named
special.test.ts
are unignored because the pattern appear after the pattern that ignore files ending with.test.js
. - Only the
test
directory at the project’s root is ignored because the pattern doesn’t start with**/
. - The linter is executed on the
src
directory at the project’s root only.
Because
includes
pattern have a different pattern format thaninclude
andignore
we made some adjustments:- We added the pattern
**
infiles.includes
to ensure that all files are included before ignoring some of them. - We added the prefix
**/
for patterns that must match at any level of the file hierarchy.
- Ignore all files ending with
-
noUndeclaredVariables
no longer reports TypeScript types.In TypeScript projects, developers often use global declaration files to declare global types. Biome is currently unable to detect these global types. This creates many false positives for
noUndeclaredVariables
.TypeScript is better suited to perform this kind of check. As proof of this, TypeScript ESLint doesn’t provide any rule that extends the
no-undef
ESLint rule.This is why Biome 1.9 introduced a new option
checkTypes
which, when it is set to false, ignores undeclared type references. The option was set totrue
by default.This option is now set to
false
by default. To get the previous behavior, you have to setcheckTypes
totrue
:{"linter": {"rules": {"correctness": {"noUndeclaredVariables": {"level": "on","options": { "checkTypes": true }}}}}} -
The rule
noUnusedVariables
no longer reports unused function parameters. UsenoUnusedFunctionParameters
. -
Fixed #5564.
noTypeOnlyImportAttributes
now ignores files ending with the extension.ts
when the type field ofpackage.json
is set tocommonjs
. -
The Biome formatter no longer adds a trailing comma in
.json
files, even whenjson.formatter.trailingCommas
is set totrue
. -
Prettier 3.4 introduced a change in their normalization process of string literals: it no longer unescapes useless escape sequences. Biome now matches the new behavior of Prettier when formatting code. This affects the JSON and JavaScript formatters.
-
Reduced accepted values for formatter options:
- The option
--quote-style
doesn’t acceptSingle
andDouble
anymore. - The option
--quote-properties
doesn’t acceptAsNeeded
andPreserve
anymore. - The option
--semicolons
doesn’t acceptAsNeeded
andAlways
anymore. - The option
--arrow-parenthesis
doesn’t acceptAsNeeded
andAlways
anymore. - The option
--trailing-commas
doesn’t acceptES5
,All
andNone
anymore. - The option
--attribute-position
doesn’t acceptSingle
andMultiline
anymore.
- The option
-
Removed the option
enumMemberCase
from the lint ruleuseNamingConvention
.enumMemberCase
is an option that allows to customize the enforced case for TypeScript’s enum members. The option was introduced prior to theconventions
option that allows to do the same thing.The following configuration…
{"linter": {"rules": {"style": {"useNamingConvention": {"level": "on","options": {"enumMemberCase": "PascalCase"}}}}}}…must be rewritten as:
{"linter": {"rules": {"style": {"useNamingConvention": {"level": "on","options": {"conventions": [{"selector": { "kind": "enumMember" },"formats": ["PascalCase"]}]}}}}}}Run
biome migrate --write
to turnenumMemberCase
intoconventions
automatically. -
Removed support for
BIOME_LOG_DIR
.The environment variable
BIOME_LOG_DIR
isn’t supported anymore.Use
BIOME_LOG_PATH
instead. -
Remove deprecated rules.
The following deprecated rules have been deleted:
noInvalidNewBuiltin
noNewSymbol
useShorthandArrayType
useSingleCaseStatement
noConsoleLog
Run the command
biome migrate --write
to update the configuration. -
Removed the deprecated
indentSize
option.The deprecated option
indentSize
, and its relative CLI options, has been removed:- Configuration file:
formatter.indentSize
- Configuration file:
javascript.formatter.indentSize
- Configuration file:
json.formatter.indentSize
- CLI option
--indent-size
- CLI option
--javascript-formatter-indent-size
- CLI option
--json-formatter-indent-size
Use
indentWidth
and its relative CLI options instead. - Configuration file:
-
Removed
ROME_BINARY
. UseBIOME_BINARY
instead. -
Removed support for legacy suppressions.
Biome used to support “legacy suppressions” that looked like this:
// biome-ignore lint(complexity/useWhile): reasonThis format is no longer supported.
-
Removed support for
max_line_length
from.editorconfig
, as it isn’t part of the official spec anymore. -
Removed support for
rome-ignore
suppression comments.Use
biome-ignore
suppression comments instead. -
Removed support for
rome.json
.Use
biome.json
orbiome.jsonc
instead. -
Removed the option
all
from the linter.The options
linter.rules.all
andlinter.rules.<group>.all
has been removed.The number of rules in Biome have increased in scope and use cases, and sometimes some of them can conflict with each other.
The option was useful at the beginning, but now it’s deemed harmful, because it can unexpected behaviours in users projects.
To automatically remove it, run the following command:
Terminal window biome migrate --write -
Removed the option
trailingComma
from the configuration and the CLI. Use the optiontrailingCommas
instead:{"javascript": {"formatter": {"trailingComma": "es5""trailingCommas": "es5"}}}biome format --trailing-comma=es5biome format --trailing-commas=es5 -
Removed
--apply
and--apply-unsafe
.The CLI options
--apply
and--apply-unasfe
aren’t accepted anymore. Use--write
and--write --unafe
instead:biome check --apply-unsafebiome check --write --unsafebiome check --applybiome check --write -
Removed support for
assert
syntax.Biome now longer supports the
assert
syntax, use the newwith
syntax insteadimport {test} from "foo.json" assert { for: "for" }export * from "mod" assert { type: "json" }import {test} from "foo.json" with { for: "for" }export * from "mod" with { type: "json" } -
Fixed #5495: The rule
noBlankTarget
has been updated to accept therel="noopener"
in addition torel="noreferrer"
. In addition, an option has been added that allowsrel="noreferrer"
to be disabled.The rule has been moved from the
a11y
group to thesecurity
group. -
The rule
useImportRestrictions
has been renamed tonoPrivateImports
, and its functionality has been significantly upgraded.Previously, the rule would assume that any direct imports from modules inside other directories should be forbidden due to their package private visibility.
The updated rule allows configuring the default visibility of exports, and recognises JSDoc comments to override this visibility. The default visibility is now
**public**
, but can be set to**package**
, or even**private**
. Refer to the documentation of the rule to understand how to leverage the JSDoc comments.noPrivateImports
is now recommended by default. -
The Biome daemon now reuses its workspace across connections. This allows multiple clients to reuse the same documents and other cached data that we extract from them.
This primarily affects our IDE extensions: If you open multiple IDEs/windows for the same project, they’ll connect to the same daemon and reuse each other’s workspace.
The Biome CLI is unaffected unless you opt in with the
--use-server
argument. -
Biome no longer treats too large files as errors.
Previously, files that exceed the configured size limit would throw an error, and the CLI would exit with an error code.
Now, the CLI ignores the file, emits an information diagnostic and doesn’t exit with an error code.
-
Change the group of some rules, promote nursery rules and update the recommended rule set.
The following rules have been moved to a new group:
- complexity/noArguments
- complexity/noCommaOperator
- complexity/noFlatMapIdentity
- complexity/noUselessContinue
- complexity/useNumericLiterals
- correctness/useValidTypeof
- performance/noNamespaceImport
- style/useArrayLiterals
- suspicious/noWith
New rules are incubated in the nursery group. Once stable, we promote them to a stable group. Use the
biome migrate
command to automatically migrate nursery rules that have been promoted.The following CSS rules have been promoted:
- correctness/noMissingVarFunction
- correctness/noUnknownPseudoClass
- correctness/noUnknownPseudoElement
- correctness/noUnknownTypeSelector
- style/noDescendingSpecificity
- style/noValueAtRule
- suspcious/noDuplicateCustomProperties
- suspcious/noDuplicateProperties
The following GraphQL rules have been promoted:
The following JavaScript rules have been promoted:
- a11y/noStaticElementInteractions
- a11y/useAriaPropsSupportedByRole(recommended)
- a11y/useValidAutocomplete
- complexity/noUselessEscapeInRegex
- complexity/noUselessStringRaw
- performance/noDynamicNamespaceImportAccess
- performance/noImgElement
- style/noCommonJs
- style/noEnum
- style/noExportedImports
- style/noHeadElement
- style/noNestedTernary
- style/noProcessEnv
- style/noRestrictedImports
- style/noRestrictedTypes
- style/noSubstr
- style/useAtIndex
- style/useCollapsedIf
- style/useComponentExportOnlyModules
- style/useConsistentCurlyBraces
- style/useConsistentMemberAccessibility
- style/useTrimStartEnd
- suspicious/noDocumentCookie
- suspicious/noDocumentImportInPage
- suspicious/noDuplicateElseIf
- suspicious/noHeadImportInDocument
- suspicious/noIrregularWhitespace
- suspicious/noOctalEscape
- suspicious/noTemplateCurlyInString
- suspicious/useAdjacentOverloadSignatures
- suspicious/useGoogleFontDisplay
- suspicious/useGuardForIn
- suspicious/useStrictMode
Moreover, the following JavaScript rules are now recommended:
- complexity/noUselessUndefinedInitialization
- complexity/useArrayLiterals
- correctness/noConstantMathMinMaxClamp
- correctness/noUnusedFunctionParameters (recommended by ESLint)
- correctness/noUnusedImports
- correctness/noUnusedPrivateClassMembers (recommended by ESLint)
- correctness/noUnusedVariables (recommended by ESLint)
- complexity/useDateNow
And the following style rules are no longer recommended:
-
Update the default severity level of lint rules.
Every diagnostic emitted by Biome has a severity level set to
error
,warn
, orinfo
. Previously, all recommended lint rules had a default severity level set toerror
. All other lint rules had a default severity level set towarn
.We have adjusted the default severity level of every rule, whether recommended or not, to better communicate the severity that a diagnostic highlights.
- Rules that report hard errors, likely erroneous code, dangerous code, or accessibility issues now have a default severity level of
error
. - Rules that report possibly erroneous codes, or code that could be cleaner if rewritten in another way now have a default severity level of
warn
. - Rules that reports stylistic suggestions now have a default severity level of
info
.
You can use the CLI option
--diagnostic-level=error
to display only errors, or--diagnostic-level=warning
to display both errors and warnings. By default, all diagnostics are shown. You can also use the CLI option--error-on-warnings
to make the command fail when warnings are emitted. - Rules that report hard errors, likely erroneous code, dangerous code, or accessibility issues now have a default severity level of
-
Reworked some recommended rules recommended to be less pedantic and blocking. This is a breaking change if your project relied on those rules to block the CI in case of violations; if that’s the case, you should raise their severity level to error.
Some rules aren’t recommended anymore, and some others return a different severity.
The following rules return a warning diagnostic:
noDelete
noForEach
noSuspiciousSemicolonInJsx
noThisInStatic
noUnusedLabels
The following rules return an information diagnostic:
noUselessCatch
noUselessConstructor
noUselessEmptyExport
noUselessFragments
noUselessLabel
noUselessLoneBlockStatements
noUselessSwitchCase
noUselessTernary
noUselessThisAlias
noUselessTypeConstraint
noFlatMapIdentity
The following rules aren’t recommended anymore:
noDelete
noForEach
The rule
noRenderReturnValue
anduseExhaustiveDependencies
are only recommended when thereact
domain is enabled. -
Renamed the global option
--skip-errors
to--skip-parse-errors
. -
Remove the code action
quickfix.suppressRule
.The code action
quickfix.suppressRule
was removed in favour of two new code actions:quickfix.suppressRule.inline.biome
: a code action that adds a suppression comment for each violation.quickfix.suppressRule.topLevel.biome
: a code action that adds a suppression comment at the top of the file which suppresses a rule for the whole file.
Given the following code
let foo = "one";debugger;The code action
quickfix.suppressRule.inline.biome
will result in the following code:// biome-ignore lint/style/useConst: <explanation>let foo = "one";// biome-ignore lint/suspicious/noDebugger: <explanation>debugger;The code action
quickfix.suppressRule.topLevel.biome
, instead, will result in the following code:/** biome-ignore lint/suspicious/noDebugger: <explanation> *//** biome-ignore lint/style/useConst: <explanation> */let foo = "one";debugger; -
Changed default formatting of
package.json
.When Biome encounters a file called
package.json
, by default it will format the file with all objects and arrays expanded.{ "name": "project", "dependencies": { "foo": "latest" } }{"projectName": "project","dependencies": {"foo": "^1.0.0"}} -
The
organizeImports
is now part of Biome Assist. -
The rule
noVar
now belongs to thesuspicious
group -
The rule
useWhile
now belongs to thecomplexity
group. -
The rule
useImportExtensions
has been updated to suggest actual file extensions instead of guesses based on hueristics.As part of this, the
suggestedExtensions
option has been removed. A simpler, new option calledforceJsExtensions
has been introduced for those who usetsc
’s"module": "node16"
setting.The rule also no longer reports diagnostics to add an extension when the path doesn’t exist at all, with or without extension.
-
Fixed #4545:
useNamingConvention
now correctly ignores declarations inside TypeScript’s external modules.The following interface name is no longer reported by the rule:
declare module "myExternalModule" {export interface my_INTERFACE {}} -
The rule
useAltText
no longer checks the element’s attributes containing object spread.The following code doesn’t trigger the rule anymore:
<img src="test.png" alt={alt} {...restProps}></img> -
The rule
useNamingConvention
no longer accepts non-ASCII characters by default.Prior to Biome 2.0, non-ASCII names were accepted by default. They are now rejected.
For example, the following code is now reported as invalid by the
useNamingConvention
rule.let johnCafé;If you want to allow non ASCII filenames and non-ASCII identifiers, you need to set the
requireAscii
options in your Biome configuration file tofalse
:{"linter": {"rules": {"style": {"useFilenamingConvention": {"level": "on","options": {"requireAscii": false}}"useFilenamingConvention": {"level": "on","options": {"requireAscii": false}}}}}} -
Renamed the rule
noUnnecessaryContinue
tonoUselessContinue
. Run the commandbiome migrate
to update your configuration. -
Renamed the rule
noMultipleSpacesInRegularExpressionLiterals
tonoAdjacentSpacesInRegex
. Run the commandbiome migrate
to update your configuration.
Minor Changes
Section titled “Minor Changes”-
An option called
allowNoReferrer
has been added to thenoBlankTarget
rule.By default,
noBlankTarget
accepts bothrel="noopener"
andrel="noreferrer"
with links that havetarget="_blank"
. This is because the latter implies the former, so either one is sufficient to mitigate the security risk.However, allowing
rel="noreferrer"
may still be undesirable, because it can break tracking, which may be an undesirable side-effect. As such, you can setallowNoReferrer: false
to only acceptrel="noopener"
. -
Added new option
javascript.parser.jsxEverywhere
. This new option allows to control whether Biome should expect JSX syntax in.js
/.mjs
/.cjs
files.When
jsxEverywhere
is set tofalse
, having JSX syntax like<div></div>
inside.js
/.mjs
/.cjs
files will result in a parsing error.Despite the name of the option, JSX is never supported inside
.ts
files. This is because TypeScript generics syntax may conflict with JSX in such files.This option defaults to
true
. -
Add a new JS assist rule -
useSortedKeys
which enforces ordering of a JS object properties. This rule will consider spread/calculated keys e.g[k]: 1
as non-sortable. Instead, whenever it encounters a non-sortable key, it will sort all the previous sortable keys up until the nearest non-sortable key, if one exist. This prevents breaking the override of certain keys using spread keys.Source: https://perfectionist.dev/rules/sort-objects
// Base// fromconst obj = {b: 1,a: 1,...g,ba: 2,ab: 1,set aab(v) {this._aab = v;},[getProp()]: 2,aba: 2,abc: 3,abb: 3,get aaa() {return "";},};// toconst obj = {a: 1,b: 1,...g,set aab(v) {this._aab = v;},ab: 1,ba: 2,[getProp()]: 2,get aaa() {return "";},aba: 2,abb: 3,abc: 3,}; -
Added the new rule
noFloatingPromises
. -
Added the new rule
noImportCycles
. -
Added the new rule
noTsIgnore
. -
Added the new rule
noUnwantedPolyfillio
. -
Added a format option
expand
for Javascript and JSON formatters. The option allows to enforce the formatting of arrays and objects on multiple lines, regardless of their length. It has three options:When set to
auto
(default), objects are expanded if the first property has a leading newline. Arrays are collapsed when they fit to a single line. For example, both styles below are considered as already formatted:const obj = {foo: "bar",};const obj = { foo: "bar" };When set to
always
, objects and arrays are always expanded.When set to
never
, objects and arrays are never expanded when they fit in a single line. It is equivalent to Prettier’s Object Wrap option withcollapse
. -
The nursery rule
noUnresolvedImports
has been added.Importing a non-existing export is an error at runtime or build time. With this rule, Biome can detect such incorrect imports and report errors for them.
Note that if you use TypeScript, you probably don’t want to use this rule, since TypeScript already performs such checks for you.
-
The rule
noFocusedTests
can now detect the usage of focused tests inside loops.// invaliddescribe.only.each([["a"], ["b"]])("%s", (a) => {});it.only.each([["a"], ["b"]])("%s", (a) => {});test.only.each([["a"], ["b"]])("%s", (a) => {});// validdescribe.each([["a"], ["b"]])("%s", (a) => {});it.each([["a"], ["b"]])("%s", (a) => {});test.each([["a"], ["b"]])("%s", (a) => {}); -
Linter groups now accept new options to enable/disable all rules that belong to a group, and control the severity of the rules that belong to those groups.
For example, you can downgrade the severity of rules that belong to
"style"
to emit"info"
diagnostics:{"linter": {"rules": {"style": "info"}}}You can also enable all rules that belong to a group using the default severity of the rule using the
"on"
option:{"linter": {"rules": {"complexity": "on"}}} -
Biome assist is a new feature of the Biome analyzer. The assist is meant to provide actions. Actions differ from linter rules in that they aren’t meant to signal errors.
The assist will provide code actions that users can opt into via configuration or via IDEs/editors, using the Language Server Protocol.
The assist is enabled by default. However, you can turn if off via configuration:
{"assist": {"enabled": false}}You can turn on the actions that you want to use in your configuration. For example, you can enable the
useSortedKeys
action like this:{"assist": {"actions": {"source": {"useSortedKeys": "on"}}}}Alternatively, IDE/editor users can decide which action to apply on save directly from the editor settings, as long as the assist is enabled.
For example, in VS Code you can apply the
useSortedKeys
action when saving a file by adding the following snippet insettings.json
:{"editor.codeActionsOnSave": {"source.biome.useSortedKeys": "explicit"}}In Zed, you can achieve the same by adding the following snippet in
~/.config/zed/settings.json
:{"code_actions_on_format": {"source.biome.useSortedKeys": true}} -
Biome migrate eslint outputs a better overriding behavior.
A Biome rule can have multiple ESLint equivalent rules. For example, useLiteralKeys has two ESLint equivalent rules: dot-notation and @typescript-eslint/dot-notation.
Previously, Biome wouldn’t always enable a Biome rule even if one of its equivalent rules was enabled. Now Biome uses the higher severity level of all the equivalent ESLint rules to set the severity level of the Biome rule.
The following ESLint configuration…
{"rules": {"@typescript-eslint/dot-notation": "error","dot-notation": "off"}}…is now migrated to…
{"linter": {"rules": {"complexity": {"useLiteralKeys": "error"}}}}…because
error
is higher thanoff
. -
Add useSymbolDescription.
-
Enhanced the command
migrate eslint
. Now the command shows which ESLint rules were migrated, and which rules aren’t supported yet../eslint.config.js migrate ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ℹ 42% (3/7) of the rules have been migrated.ℹ Migrated rules:- getter-return- prefer-const- @typescript-eslint/require-awaitℹ Rules that can be migrated to an inspired rule using --include-inspired:- @typescript-eslint/parameter-propertiesℹ Rules that can be migrated to a nursery rule using --include-nursery:- @typescript-eslint/switch-exhaustiveness-checkℹ Stylistic rules that the formatter may support (manual migration required):- semiℹ Unsupported rules:- block-scoped-varconfiguration ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ℹ Migration results:- ./biome.json: configuration successfully migrated. -
Suppression of syntax rules
Added support for suppressing syntax rules. Syntax rules are particular rules meant to complement the parser, hence they can’t be configured.
Biome now allows to suppress those rules. This can, for example, be useful in case the rule is affected by a bug. However, this is more an escape hatch, so if a syntax rule requires a suppression, please file an issue.
Example:
// biome-ignore syntax/correctness/noTypeOnlyImportAttributes: bugimport type { MyType } from "my-esm-pkg" with { "resolution-mode": "import" };Biome now requires all
biome-ignore-start
suppressions to have an equivalentbiome-ignore-end
comment. -
Add a new lint rule
noConstantBinaryExpression
. This rule is inspired from ESLint’s no-constant-binary-expression rule. -
The CLI options
--only
and--skip
now accept rule and action names without prefixing the group name.Previously
--only=noDebugger
was rejected. You had to add the group name:--only=suspicious/noDebugger
. -
Fixed #3574:
noUnusedImports
now reports empty named imports and suggests their removal.The rule now suggests the removal of empty named imports such as:
import {} from "mod"; -
Added the new rule
useAdjacentGetterSetter
, which enforces getters and setters for the same property to be adjacent in class and object definitions.Example (Invalid): Name getter and setter are not adjacent:
class User {get name() {return this._name;}constructor() {}set name(value) {this._name = value;}}**Example (Invalid): Getter should go before the setter.
const user = {set name(value) {this._name = value;},get name() {return this._name;},};Example (Valid): Name getter and setter are adjacent:
class User {get name() {return this._name;}set name(value) {this._name = value;}get age() {return this._age;}set age(age) {this._age = age;}} -
Added new rule useConsistentResponse which suggests to use static Response.json() and Response.redirect() methods instead of
new Response
when possible.Example:
new Response(JSON.stringify({ value: 1 }));Response.json({ value: 1 }); -
Biome users can now configure code actions from linter rules as well as assist actions directly in the settings of their IDE/editor.
For example, let’s consider the lint rule
noSwitchDeclarations
, which has an unsafe fix. Previously, if you wanted to use this rule, you were “forced” to enable it via configuration, and if you wanted to apply its fix when you saved a file, you were forced to mark the fix as safe:{"linter": {"rules": {"correctness": {"noSwitchDeclarations": {"level": "error","fix": "safe"}}}}}Now, you can benefit from the code action without making the fix safe for the entire project. IDEs and editors that are LSP compatible allow to list a series of “filters” or code actions that can be applied on save. In the case of VS Code, you will need to add the following snippet in the
settings.json
:{"editor.codeActionsOnSave": {"quickfix.biome.correctness.noSwitchDeclarations": "explicit"}}Upon save, Biome will inform the editor the apply the code action of the rule
noSwitchDeclarations
. -
Fixed #3401:
noUnusedImports
now keeps comments separated from the import with a blank line.For example:
// Orphan comment// Header commentimport {} from "mod"; -
Added a new
propertyAssignment
option to thenoParameterAssign
rule. This option allows to configure whether property assignments on function parameters are permitted. By default,propertyAssignment
is set toallow
. Setting it todeny
enforces stricter immutability by disallowing property mutations on function parameters. -
The formatter option
bracketSpacing
is now also supported in JSON files. -
useValidTypeof
now accepts comparisons with variables.Previously, the rule required to compare a
typeof
expression against anothertypeof
expression or a valid string literal. We now accept more cases, notably comparison against a variable:if (typeof foo === bar) {// ...} -
Added the new rule
noNestedComponentDefinitions
, which disallows nested component definitions in React components.This rule is useful for preventing potential performance issues and improving code readability by ensuring that components are defined at the top level.
Example (Invalid):
function ParentComponent() {function ChildComponent() {return <div>Hello</div>;}return <ChildComponent />;}Example (Valid):
function ChildComponent() {return <div>Hello</div>;}function ParentComponent() {return <ChildComponent />;} -
Added the new rule
noDestructuredProps
, which disallow the use of destructured props in Solid projects. -
You can now enable lint rules using the default severity suggested by Biome using the new variant
"on"
, when enabling a rule.For example, the default severity of the rule
style.noVar
iserror
, so you would use"on"
, and then linting a code that usesvar
, will result in an error:{"linter": {"recommended": false,"rules": {"style": {"noVar": "on"}}}}main.js var name = "tobias";The command
biome lint main.js
will result in an error due to the default severity assigned tonoVar
.Refer to the documentation page of each rule to know their suggested diagnostic severity, or use the command
biome explain <RULE_NAME>
:Terminal window biome explain noVar -
Biome VCS integration now supports nested ignore files.
For
git
, if a.gitignore
is found in a nested folderroot/packages/foo/
, and it contains the patterndist/
, only files and directories insideroot/packages/foo/dist
are matched. -
Added the rule useUniqueElementIds. This rule disallows the use of static IDs in React components. It encourages to generate unique IDs for accessibility purposes using
useId
.The following code is now reported as invalid:
function App() {return <div id="static-id" />;}The following code is now reported as valid:
import { useId } from "react";function App() {const id = useId();return <div id={id} />;} -
Added the new JavaScript rule
useConsistentObjectDefinition
rule. The rule enforces a consistent style for the definition of objects:By default, the rule enforces a shorthand style:
const validShorthand = {// Property shorthandfoo,// Method shorthandmethod() {return "method";},};Alternatively, the rule can be configured to enforce an explicit style:
const invalidExplicit = {// Basic property shorthand violationsfoo: foo,// Method shorthand violationsmethod: function () {return "method";},}; -
Introduced more advanced logging capabilities:
Every Biome CLI command can now be passed a
--log-file=<path>
argument, which will write all log messages for that invocation to the given path instead ofstdout
.In addition, the
--log-level
parameter now also accepts atracing
value. When--log-level=tracing
is used, Biome also prints timing information from tracing spans to the log.Combined with Biome’s ability to print logs in JSON format, and the
jq
command line utility, this allows you to perform advanced analysis on Biome’s internal performance.For example, if you want to figure out which paths take the longest when building the module graph, you can use the following commands:
Terminal window biome lint --log-level=tracing --log-kind=json --log-file=tracing.jsoncat tracing.json | jq '. | select(.span.name == "update_module_graph") | { path: .span.path, time_busy: .["time.busy"], time_idle: .["time.idle"] }' > filtered.jsonNow you will have a file called
filtered.json
with all the relevant timings, together with the paths used during the invocations. -
Added options to
suspicious/noConfusingLabels
to allow specific labels. -
Fixed #4549: noUnknownProperty now accepts more known CSS properties.
['anchor-default', 'anchor-scroll', 'inset-area', 'position-animation', 'position-fallback', 'position-fallback-bounds', 'position-try-options']['anchor-scope', 'interpolate-size', 'line-fit-edge', 'masonry', 'masonry-auto-tracks', 'masonry-direction', 'masonry-fill', 'masonry-flow', 'masonry-slack', 'masonry-template-areas', 'masonry-template-tracks', 'position-anchor', 'position-area', 'position-try-fallbacks', 'position-visibility', 'scroll-start-target', 'text-box', 'view-transition-class', 'view-transition-group']This change replaces deprecated properties, improving CSS validation.
-
LSP clients can now override the configuration path for each workspace, by responding to
workspace/configuration
requests. -
Added the new CSS rule
noImportantStyles
, which prevents the use of!important
inside CSS declarations. -
Biome now emits a warning diagnostic if the configuration contains an out-of-sync schema URL.
-
Introduced a new configuration setting
files.experimentalScannerIgnores
.This setting may be used to configure a set of file and folder names that should be unconditionally ignored by Biome’s scanner.
Biome maintains an internal list of default ignore entries, which is based on user feedback and which may change in any release. This setting allows overriding this internal list completely.
This is considered an advanced feature that users should not need to tweak themselves, but they can as a last resort. This setting can only be configured in root configurations, and is ignored in nested configs.
Entries must be file or folder names. Specific paths and globs are not supported.
Examples where this may be useful:
{"files": {"experimentalScannerIgnores": [// You almost certainly don't want to scan your `.git` folder, which// is why it's already ignored by default:".git",// But the scanner does scan `node_modules` by default. If you// *really* don't want this, you can ignore it like this:"node_modules",// But it's probably better to ignore a specific dependency.// For instance, one that happens to be particularly slow to scan:"RedisCommander.d.ts",],},}Please be aware that rules relying on the module graph or type inference information may be negatively affected if dependencies of your project aren’t (fully) scanned.
-
Added the new rule useSingleJsDocAsterisk which enforces JSDoc comment lines to start with a single asterisk.
// Invalid/**** Description*/// Valid/*** Description*/ -
The CLI flag
--javascript-attribute-position
was renamed to--javascript-formatter-attribute-position
for consistency. -
Introduced the
domains
linter feature. The Biome linter now has a new way to opt-in rules, with a concept calleddomains
.Domains can be seen as concepts shared by different rules.
You can enable and disable multiple rules that belong to a domain. When you assign
"all"
, Biome will enable all the rules, when you assign"none"
, Biome will disable the rules, when you assign “recommended”, Biome will enable all rules of the domain that are recommended.biome.jsonc {linter: {domains: {test: "all", // all rules that belong to this domain are enabledreact: "recommended", // only the recommended rules from this domain are enabledsolid: "none", // rules related to Solid are disabled},},}New domains introduced:
test
: it will enable rules:noExportsInTest
noExcessiveNestedTestSuites
noDuplicateTestHooks
noFocusedTests
And it will inject the following globals:after
afterAll
afterEach
before
beforeEach
beforeAll
describe
it
expect
test
next
: it will enable rules for Next.js projects:useExhaustiveDependencies
useHookAtTopLevel
noImgElement
noHeadImportInDocument
noHeadImportInDocument
react
: it will enable rules for React projects:useExhaustiveDependencies
useHookAtTopLevel
solid
: it will enable rules for Solid projects:noReactSpecificProps
For more information regarding how Biome enables rules via domains, please refer to the documentation page of each rule.
-
Biome now prints diagnostics sorted by their severity. The order is the following:
- information
- warning
- error
This means that error diagnostics are printed last, so users can see them first.
-
Added the new rule
useIterableCallbackReturn
, which enforces consistent return values in iterable callbacks.The following methods require a return value in their callback:
every
filter
find
findIndex
findLast
findLastIndex
flatMap
map
reduce
reduceRight
some
sort
toSorted
—from
(when called onArray
)
The rule disallows a return value inside the callback of the method
forEach
.Examples:
[].map(() => {// Missing return value});[].forEach(() => {return 1; // Disallowed}); -
Added the new rule
noReactPropAssign
, based on the react-hooks rule react-hooks/react-compilerThe following code is now reported as invalid:
function Foo(props) {props.bar = `Hello ${props.bar}`;return <div>{props.bar}</div>;}The following code is now reported as valid:
function Foo({ bar }) {bar = `Hello ${bar}`;return <div>{bar}</div>;} -
Added new rule
noBitwiseOperators
, which disallows bitwise operators. -
The Biome analyzer now supports a new top-level suppression. These suppression have to be placed at the top of the file, and they must be followed by two newlines (
\n\n\
).The analyzer rules specified inside the block comment will be suppressed for the whole file.
In the example, we suppress the rules
lint/style/useConst
andlint/suspicious/noDebugger
for the whole file:main.js /*** biome-ignore-all lint/style/useConst: i like let* biome-ignore-all lint/suspicious/noDebugger: needed now*/let path = "/path";let _tmp = undefined;debugger;In this other example, we suppress
lint/suspicious/noEmptyBlock
for a whole CSS file:/**/* biome-ignore-all lint/suspicious/noEmptyBlock: it's fine to have empty blocks*/a {}span {}A new diagnostic is emitted if
biome-ignore-all
suppression isn’t placed at the top of the file:file.js:3:1 suppressions/incorrect ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━! Top level suppressions can only be used at the beginning of the file.2 │ let foo = 2;> 3 │ /**│ ^^^> 4 │ * biome-ignore-all lint/style/useConst: reason> 5 │ */│ ^^6 │ let bar = 33;i Rename this to biome-ignore2 │ let foo = 2;3 │ /**> 4 │ * biome-ignore-all lint/style/useConst: reason│ ^^^^^^^^^^^^^^^^5 │ */6 │ let bar = 33; -
Added the new rule
useNumericSeparators
, which encourages the use of numeric separators to improve readability. -
useImportExtensions now checks imports with sub extensions.
- import 'styles.css'+ import 'styles.css.ts' -
It’s possible to override the option
files.maxSize
. This option is helpful if you need to process specific files that exceed the defaultmaxSize
:{"overrides": [{"includes": ["dist/**"]"files": {"maxSize": 20000}}]} -
Added the new CLI option called
--threads
to theci
command. It allows to control the numbers of threads that can be used when using the Biome CLI.It’s possible to use the environment variable
BIOME_THREADS
as an alternatives.This feature is useful when running the CLI in environments that have limited resources, for example CI/CD.
Terminal window biome ci --threads=1BIOME_THREADS=1 biome ci -
Added the new rule
useObjectSpread
, which prefers object spread syntax overObject.assign()
when constructing new objects.Example (Invalid): Using Object.assign with an empty object:
Object.assign({}, foo);Object.assign({}, { foo: "bar" });Example (Invalid): Using Object.assign with object literal as first argument:
Object.assign({ foo: "bar" }, baz);Object.assign({}, baz, { foo: "bar" });Example (Valid): Using object spread syntax:
({ ...foo });({ ...baz, foo: "bar" });Example (Valid): Modifying existing objects is allowed:
Object.assign(foo, { bar: baz });Object.assign(foo, bar, baz); -
Added an option to the
lint
command called--suppress
. The new option suppresses a violation instead of applying a rule fix. The option accepts a string that is used as reason of the suppression comment.When running the following command, it will add the suppression comment:
Terminal window biome lint --write --suppress="Migration to Biome"debugger;foo == bar;// biome-ignore lint/suspicious/noDebugger: Migration to Biomedebugger;// biome-ignore lint/suspicious/noDoubleEquals: Migration to Biomefoo == bar; -
Add an
ignoreRestSiblings
option intonoUnusedVariables
.When this option is set to
false
, the rule will not ignore variables that created using the rest pattern:{"linter": {"rules": {"correctness": {"noUnusedVariables": {"level": "error","options": {"ignoreRestSiblings": false}}}}}}const { lorem, ...test } = bar; // the variable "test" will trigger the ruleconsole.log(lorem); -
Upgraded some unsafe fixes to safe fixes.
The following rules have now a safe fix:
-
Added support for monorepos. The feature will work out of the box for the majority of the users. If your project has nested configuration files, use the command
biome migrate
from the root of the project.Monorepo support in Biome is done in a single way. Create a
biome.json
at the root of the project. This configuration file is now called the root configuration. Then, each nested configuration file must specify the new field"root": false
.We also introduced a new microsyntax for extending a nested configuration from the root configuration, which is
"extends": "//"
. This new syntax means “this config extends from the root config”. When using this microsyntax, you may omit the"root": false
field as it is implied.Note that nested configs are not required to extend from the root config, and you can still have independent nested configs, as well as nested configs that extend from other files. In those cases,
"root": false
must be specified explicitly. -
Added support for formatting
.html
files. The formatting is considered experimental, and it’s only opt-in via configuration:{"html": {"formatter": {"enabled": true}}}Biome formatter attempts to format as Prettier, however some default options might differ.
An option
html.formatter.selfCloseVoidElements
allows to control whether the trailing/
of void elements should be printed.By default, Biome formatter will remove the
/
:<input /><input>If you come from Prettier and you want to keep the same formatting behaviour, you should set the option to
"always"
:{"html": {"formatter": {"selfCloseVoidElements": "always"}}}<input><input />Use to the command
biome migrate prettier
to apply this change automatically. -
Added an unsafe fix to the rule
useExhaustiveDependencies
.For example, this violation will provide the following code fix:
import { useEffect } from "react";function MyComponent() {let a = 1;useEffect(() => {}, [a]);}× This hook specifies more dependencies than necessary: a3 │ function MyComponent() {4 │ let a = 1;> 5 │ useEffect(() => {}, [a]);│ ^^^^^^^^^6 │ }7 │i This dependency can be removed from the list.3 │ function MyComponent() {4 │ let a = 1;> 5 │ useEffect(() => {}, [a]);│ ^6 │ }7 │i Unsafe fix: Remove the extra dependencies from the list.5 │ ··useEffect(()·=>·{},·[a]);│ - -
The rule
useExhaustiveDependencies
now reports a diagnostic when the dependency list is not an array literal. -
Added the new rule
useIndexOf
, based on the unicorn rule prefer-array-index-of -
Added a new rule
useExhaustiveSwitchCases
, which detects any missing cases for switch statements. Currently, it supports only literal union types.For example:
type Day =| "Monday"| "Tuesday"| "Wednesday"| "Thursday"| "Friday"| "Saturday"| "Sunday";const day: Day = "Monday";let result = 0;switch (day) {case "Monday": {result = 1;break;}}The switch statement is missing other cases than
'Monday'
, which will cause a runtime error. To fix this issue, add missing cases or a default case to the statement. -
Fixed #4416: The rules
useExportType
anduseImportType
now ignore TypeScript declaration files. -
Added the new rule
useForComponent
.This rule enforces usage of Solid’s
<For />
component for mapping an array to JSX elements. -
Added new lint rule
noShadow
, a port of eslint’sno-shadow
.This rule disallows variable declarations from shadowing variables declared in an outer scope. For example:
const foo = 1;function bar() {const foo = 2; // This variable shadows the outer foo} -
Add
style
option for the useImportType rule.The rule now allows enforcing an import style for importing types. See the rule documentation for more details.
-
Added the new rule
useJsonImportAttribute
to enforce the use of import attributes for JSON modules.This rule ensures that all imports of
.json
files include thewith { type: "json" }
assertion, which is required to inform the JavaScript runtime that the imported file should be parsed as JSON.import jsonData from './data.json';import jsonData from './data.json' with { type: "json" };import jsonData from './data.json' with { someOtherAttribute: "value" };import jsonData from './data.json' with { type: "json", someOtherAttribute: "value" };This rule is based on the proposal in issue #6043.
-
useLiteralKeys now handles numeric keys and is declared as being the same rule as the ESLint no-useless-computed-key rule.
-
useNamingConmvention
now ignores unused variables prefixed with an underscore_
.This avoids conflicts with the unsafe fix of
noUnusedVariables
. The following code is now accepted because the variable is unused and prefixed with an underscore.const _Unknown_Style = 0; -
The package now requires
v2
of the WebAssembly packages. The internal APIs of Workspace are nowcamelCase
. -
The rule useArrayLiterals now provides a code fix.
const xs = new Array();const xs = [];The code fix is currently marked as unsafe. We plan to make it safe in a future release of Biome.
-
The command
migrate
is now able to migrate nested configuration files. -
Added the new rule
noRestrictedElements
, which prevents use of the specified HTML elements and components. -
Added the new lint rule
noAwaitInLoop
.
Patch Changes
Section titled “Patch Changes”-
Fix #5001, where the CSS formatter removes whitespace from selector preceded by a comment
-
Fixed noUnusedVariables rule to suggest unsafe fix for unused function declarations.
-
Fixed a false positive of
noUselessEscapeInRegex
where\k
was reported as useless in non-Unicode regular expressions. -
Fixed an issue where the ordering of the diagnostics wasn’t predictable.
-
Fixed a bug where the environment variable
BIOME_CONFIG_PATH
wasn’t correctly picked up. -
Biome logs a warning in case a folder contains
biome.json
andbiome.jsonc
, and it will usebiome.json
by default. -
noUndeclaredVariables is now able to bind read of value to a type-only import in ambient contexts (#4526).
In the following code,
A
is now correctly bound to the type-only import. Previously,A
was reported as an undeclared variable.import type { A } from "mod";declare class B extends A {} -
Fix #4317, setter parameter can contain a trailing comma, the following example will now parsed correctly:
export class DummyClass {set input(value: string) {}} -
Fix #4575, don’t wrap selector indentation after css comments.
-
Fix #4258, where fixed css parse error with @-moz-document url-prefix().
-
Fixed #4391: Some files from the
.vscode
directory are no longer incorrectly parsed as JSON. -
The
biome format
command now correctly handles the--skip-errors
option, allowing it to skip files with syntax errors and continue formatting the remaining valid files. When this option is used, skipped syntax errors are reported as information, since the user is already aware of them. -
biome migrate eslint
now correctly resolves the scoped package namedeslint-config
. -
Fixed #3836: The CSS parser will now correctly parse the following:
.foo {color: red;} -
Fixed a bug where the related diagnostics attached to the main diagnostics didn’t have a correct message.
-
Fixed
noAccumulatingSpread
not reporting calls toObject.assign
. The following code will now be reported:let a = [{ a: 1 }, { b: 2 }];a.reduce((acc, val) => Object.assign(acc, val), []); -
The
summary
reporter doesn’t take--max-diagnostics
into account anymore. -
Fixed #4553:
noUselessFragments
will now correctly fix JSX attributes:<Suspensefallback={<><span>Loading...</span></>}>{children}</Suspense>becomes:
<Suspense fallback={<span>Loading...</span>}>{children}</Suspense> -
Fixed #4528:
biome migrate eslint
now correctly handles shared ESLint configuration that don’t follow the ESLint naming convention.ESLint recommends that a package that exports a shared configuration be prefixed with
eslint-config-
or simply namedeslint-config
. This is only a recommendation. Packages that export shared configurations can have arbitrary names. Biome is now able to load any package. -
Fixed #4993:
useAwait
now correctly warn on functions with decorator with callback argument. -
Fixed #4756:
noDuplicateProperties
now throws lint errors properly when we use@supports
. -
Fixed #5981, where
noUnknownPseudoClass
didn’t take:global
into consideration whencssModules
is enabled. -
Fixed #2406: Biome longer expands properties of object type annotations in the only function parameter to align with Prettier.
-
Fixed #4740:
biome migrate eslint
now correctly handles ESLint configuration withnull
values in file lists. -
Fixed #4202: Align with Prettier in formatting test functions.
-
Fixed #342: The JavaScript parser now properly handles unterminated string literals, such as:
function Comp() {return (<a rel=" -
Fixed a bug where syntax rules didn’t provide an automatic way to suppress the rule. Now the LSP will show supression actions if a syntax rule is violated.
-
Fixed a CSS parser error:
@-moz-document url-prefix(https://example.com)
and@-moz-document domain(example.com)
are now valid. -
Fixed #4967: The fix for
useArrowFunction
no longer breaks function bodies starting with{
. -
Fixed #5998. The rule
noUnknownPseudoElement
now correctly cheks names of pseudo-element functions. -
Fixed #5024: Added
useJsxKeyInIterable
rule to React domain. -
Fixed #5410. Biome now correctly parse an
.editorconfig
that includes character classes in glob patterns. -
Fixed #2260: The LSP server now returns correct text edits for the specified range in
textDocument/rangeFormatting
andtextDocument/onTypeFormatting
requests. -
noUnusedVariables no longer reports top-level variables in a global declaration file as unused.
-
Type exports now support renaming types to
default
.The following code is now parsed successfully:
export { type A as default } from "./b.ts"; -
Added proper support for arrow functions in the lint rule https://biomejs.dev/linter/rules/use-explicit-type/
-
The rule
useNamingConvention
no longer suggests renaming top-level variables in a global declaration file. -
Improved context in error messages when migrating Prettier configurations
-
Allowed single spaces in
useConsistentCurlyBraces
rule. -
Fixed #4413: The GraphQL formatter no longer adds a new line at the start of block comments on Windows.
-
Fixed #5407. Now the
noUnusedImports
code fix correctly keeps top-level comments that were attached to lone imports. -
Fixed #3859: the
--skip-parse-errors
option is now applied to commands:lint
,check
, andci
. -
The
rage
command now prints the configuration path relative to the working directory, if applicable. -
Fixed #5606: We now correctly handle
.mjs
extensions in Node.js packages with"type": "commonjs"
. -
Fixed #1597:
useExhaustiveDependencies
no longer gets confused about the stability of dependencies by parentheses or type assertions. -
Fixed #4751 by checking fragments inside
JSXElement
and conditional expressions.For example, the following two cases will now be reported:
<section><><div /><div /></></section>showFullName ? <>{fullName}</> : <>{firstName}</>; -
The rule
noFallthroughSwitchCase
no longer panics on some incomplete code snippets. -
Fixed #5007: Resolved false positives in
noMissingVarFunction
forcontainer-name
. -
Fixed #4841: Shebang and top leading comments in
.cjs
files are now handled correctlyExample: shebang only (keep it as is)
#!/usr/bin/env nodeExample: comments only (keep it as is)
// commentExample: with shebang
#!/usr/bin/env node"use strict";#!/usr/bin/env node"use strict";let some_variable = "some value";Example: with comment
// comment"use strict"; // comment"use strict";// commentlet some_variable = "some value";Example: with shebang and comment
#!/usr/bin/env node"use strict";// comment#!/usr/bin/env node"use strict";// commentlet some_variable = "some value"; -
Fixes
useExhaustiveDependencies
missing dependencies being defined after the hook itself failure.Example:
import { useState, useEffect } from "react";function MyComponent() {useEffect(() => {console.log(a);}, []);let a = 1;} -
Fixed #4714: Suppression comments no longer fail on functions that themselves contain suppression comments.
This now works correctly:
// biome-ignore lint/complexity/useArrowFunction: this suppression now worksconst foo0 = function (bar: string) {// biome-ignore lint/style/noParameterAssign: even if there are other suppressions insidebar = "baz";}; -
Add @vitest/eslint-plugin to list of Biome rule sources
-
Fixed
useHookAtTopLevel
rule to properly detect React components wrapped inmemo
andforwardRef
, and correctly handle property accessors in control flow analysis.The rule now correctly identifies hooks in components like:
const TestMemo = memo(forwardRef((props, ref) => {useEffect(() => {const [test, setTest] = useState(1); // now properly flagged}, []);return <div ref={ref}>test</div>;}),);And properly handles property accessors:
function ReactComponent() {const testObj = {get print() {return "hello"; // no longer considered component return},};const callback = useCallback(() => {}, []);return <></>;} -
noMisleadingCharacterClass no longer panics on malformed escape sequences that end with a multi-byte character (#4587).
-
Fixed the flag
--bracket-spacing
that was duplicated between the global configuration and the language-specific override for JavaScript. -
Fixed #4715: The
useJsxKeyInIterable
rule now reports missing keys insideswitch
andif
statements.const data = [{ value: "a", type: "string" },{ value: 9, type: "number" },{ value: "c", type: "string" },];const MyComponent = () => {return (<>{/* if statements */}{data.map((x) => {if (x.type === "string") {return <div>{x.value}</div>; // no key, emits diagnostic} else {return <div>{x.value}</div>; // no key, emits diagnostic}})}{/* switch statements */}{data.map((x) => {switch (x.type) {case "string":return <div>{x.value}</div>; // no key, emits diagnosticcase "number":return <div>{x.value}</div>; // no key, emits diagnosticdefault:return <div key={x.value}>{x.value}</div>;}})}</>);}; -
Fixed #4121: The CSS formatter no longer indents a selector when it has leading comments.
-
Fixed an issue where react lint rules could panic Biome when some incorrect code was analyzed.
-
Fixed #4982: the JavaScript parser now throws a syntax error for the following code:
type T = import;type U = typeof import; -
Fixed a bug with the
--verbose
CLI flag. Now the printed paths are relative to the working directory. -
Fixed
noNoninteractiveElementToInteractiveRole
mistakenly flagging<li role="treeitem">
, -
Fixed #4622: Our JavaScript parser can now gracefully handle situations where we detect the parser to have stalled.
This means we don’t fail with an assertion anymore, but invalid code can trigger a regular diagnostic in such cases.
-
Fixed #342: The JavaScript parser now correctly handles invalid object member names, such as:
({params: { [paramName: string]: number } = {}}) -
Fixed #6211: previously the import organizer emitted broken code when it merged an import at the start of the file with another import and placed the merged result after a third import.
The following code is now correctly organized:
import { B } from "bc";import { C } from "bc";import { A } from "a";import { B, C } from "bc"; -
Fixed #4334: The formatter no longer inserts trailing a comma inside dynamic
import
expressions. -
Fixed #5629: useHookAtTopLevel no longer report false-positives where the hook is at the top-level in a class method.
-
Fixed #5900:
biome migrate eslint
now support a nestedfiles
property in ESLint flat configs. -
Fixed #3895: noUnusedImports no longer reports used values imported as types in an external module.
-
Fixed a case where the code fix for
noUselessFragments
would remove more than just the fragment. -
Fixed #5919. Now Biome correctly loads the configuration passed via
--config-path
when its path starts with./
e.g.--confi-path=./project/biome.json
-
Fixed #5031: CSS formatting has been improved for numbers:
.class {padding: .5em;marding: 1.0;padding: 0.5em;marding: 1;} -
Fixed #5989 where large octal escape sequences led to an overflow.
-
Implement improved error handling for the supports at rule
-
Fix #5053, now the rule correctly handles
console.log
inside arrow function expressions. -
Fix #6105: css lint rules
useSortedProperties
should skip unknown properties. -
Fixed #3229: Made formatting of compound selectors more consistent.
-
Fixed a bug where passing
--max-diagnostics=0
would return a zero code even when errors were emitted. -
Fixed a bug where Biome didn’t report any error when
--stdin-file-path
didn’t have any extension. Now Biome returns an error if--stdin-file-path
doesn’t have an extension. -
Fixed #5601: The
useSortedClasses
rule now properly preserves the original JSX quote style when sorting utility classes, preventing syntax errors. -
The fix for
useSelfClosingElements
was marked as safe and the error message was improved. -
Fixed overrides that include language-specific settings from having an effect for some languages
-
Fixed #6144: noUnusedImports reported incorrectly imports that were used as the type of parameters with the same name. In the following code, the import
name
was reported as unused.import name from "mod";function f(name: name.Readable): void {} -
The lint rules
useNamingConvention
anduseFilenamingConvention
now accept character escapes at the start of a regex group.Both these rules provide options that allow matching names against a regular expression. Previously, an escaped character at the start of a regex group reported an error. They are now accepted.
For example, the following configuration is now valid doesn’t emit an error anymore.
{"linter": {"rules": {"style": {"useNamingConvention": {"level": "on","options": {"conventions": [{"selector": {"kind": "let"},"match": "(\\n.*)"}]}}}}}} -
Fixed #5617: noDuplicateObjectKeys now transfers the leading comments of the removed member.
-
Fixed #5409: noParameterAssign now reports reassigned parameter of unparenthesized arrow functions.
The following code is now reported as invalid.
const f = (param) => {param = {}; // Reassigning a function parameter is confusing.}; -
Fixed #4875: Relative file paths are now clickable in the Jetbrains IDE terminal.
-
Fixed #4719:
bracketSameLine
now performs as expected when a comment is placed before the last JSX attribute. -
Fixed #4564: Biome no longer panics when a multi-byte character is found in a unicode escape sequence.
-
Fixed #4950: Resolved a false positive of character class range operators in regular expressions.
-
Fixed handling of top-level variables by
useExplicitType
rule (#5932). Biome now allows all variables with explicit annotations, as well as variables with trivial RHS. Biome no longer emits duplicated errors when an untyped function is assigned to an untyped variable. -
Fixed #4947: The
useTemplate
lint rule now ignores concatenated literals folded to multiple lines. -
Fixed #4568: Broken import statements no longer can cause a panic in
useExhaustiveDependencies
. -
Fixed #6042:
noUselessEscapeInString
now reports useless escapes after skipping ${ in template literals. -
Fixed #6229 where the fix of
noUnusedImports
emitted an invalid syntax. Now the following case emits a code fix that is syntactically correct:import Used, { NotUsed } from "foo";Used(); -
Fix #5682: Object patterns with a nested assignment pattern no longer break properties.
For example, the following code:
const { foo: { bar } = { bar: false } } = props;is used to be formatted into:
const { foo: { bar } = { bar: false } } = props;, while Prettier does not expand properties in this case.
-
Fixed #5620, noConsole rule now correctly handles indirect
console.log
calls and references. -
When pulling code actions from the LSP, now the first choice suggested by the client will be the safe fix.
-
Fixed #6022, now the rule
noDuplicateProperties
does’t trigger properties defined inside the@keyframes
at rule -
Enhanced the error message of the diagnostics emitted when Biome can’t parse a suppression comment.
-
Fixed link to the docs inside CLI markup
-
Fixed a bug where a suppression comment with an empty explanation was valid.
Now a suppression comment
// biome-ignore lint:
will raise a warning diagnostic. -
Fixed #4026: Comments in
grid-template
are no longer moved by the formatter. -
Fixed #3394: Resolved a false positive in
useSortedClasses
. -
Fixed #342 and #4562: Biome no longer crashes when a
declare
statement is followed by an unexpected token. -
Fixed false positive in the rule
noUnknownFunction
where thetech
function was incorrectly flagged as an unknown function. -
Fixed #4511: noLabelWithoutControl now detects
<button>
tags as input. -
Fixed #6039:
noUselessEscapeInString
no longer reports\${
escape in template literals. -
Fixed #5985, which caused the import organizer to fail the merging of a default import with a named import. The following code is now correctly organized:
import moment from 'moment';import { Moment } from 'moment';import moment, { Moment } from 'moment'; -
Fixed an issue where the
explain
command didn’t the diagnostic category when a rule was explained. -
Improved the diagnostic of the rule
noUnusedVariables
. The rule message now provides the name of the unused binding. -
Added
RegExpStringIterator
to the analyzer globals. -
Fixed #4208: noUselessFragments now handles
JsxAttributeInitializerClause
, ensuring that fragments inside expressions like<A b=<></> />
are preserved. -
Fixed #4533:
noUnknownPseudoClass
no longer reports pseudo classes after a webkit scrollbar pseudo element.The following code will no longer report a diagnostic:
::-webkit-scrollbar-thumb:hover {} -
Updates the
useJsxKeyInIterable
rule to more closely match the behavior of the ESLint plugin (e.g. mark the whole fragment as incorrect when no key is present). This also adds the option to check shorthand fragments (<></>
) -
Renamed the rule
noDuplicatedFields
tonoDuplicateFields
. Run the commandbiome migrate
to update your configuration. -
Fixed an issue where ignored files were incorrectly tracked by the Daemon.
-
Fixed #5116: noUnknownPseudoElement now supports
::slotted
. -
Fixed #5979:
biome search
now correctly skips files that don’t match the pattern’s target language. -
Fixed #4323: Fixed the case where
useSemanticElement
accidentally showed recommendations forrole="searchbox"
instead ofrole="search"
. -
Support setting
indent_size
totab
in.editorconfig
, the following config will not cause error:root = true[*]indent_size = tab -
Fixed #4565: noControlCharactersInRegex no longer panics when it encounters an unterminated unicode escape sequence.
-
Fixed #5770, Biome’s configuration file is now respected by the
migrate
command during migration -
Fixed an issue where the lexer didn’t report errors for unterminated regex or string literals, such as the following cases:
"string'str/\\217483 -
The
useKeyWithClickEvents
rule has been improved with better support for ARIA roles.Key improvements:
- Accessibility checks:
Now the rule correctly handles the following cases:
- If an element is hidden from screen readers
- If an element has the presentation role
- If an element is interactive
// No errors<div aria-hidden="true" onClick={() => {}} /> // hidden from screen reader<div role="presentation" onClick={() => {}} /> // presentation role<button onClick={() => {}} /> // interactive roleThis change ensures the rule is more accurate and helpful.
- Checks spread syntax:
Spread syntax used to be ignored, but has been changed to be pointed out for more stringent checking.
// Errors<div {...props} onClick={() => {}} />// No errors<div {...props} onClick={() => {}} onKeyDown={foo} />;- Refactor:
Now the rule uses the aria roles to determine if an element is interactive.
The changes shown here are meant to be closer to the original jsx-eslint’s
click-events-have-key-events
rule. -
Fixed #6029: A new line before the semicolon in the previous statement is now kept after formatting.
For example, the following code:
const foo = 3;[1, 2, 3].map((x) => x * 2);when
javascript.formatter.semicolons
isalways
, it becomes:const foo = 3;[1, 2, 3].map((x) => x * 2);when
javascript.formatter.semicolons
isasNeeded
, the original code is considered as already formatted. -
useArrayLiterals now reports all expressions using the
Array
constructors.Previously, the rule reported only use of the
Array
constructor in expressions statements.// This was reportednew Array();// This was not reportedconst xs = new Array(); -
Improved error handling for the container at-rule.
-
Fixed #4665: the LSP previously identified
.cjs
files as ESM files, making rules likenoRedundantUseStrict
reports incorrectly valid"use strict"
directives. -
Fixed #5382:
useExportType
no longer reports an identifier that bound by both a variable and a type. -
Fixed #5826:
useNumericSeparators
no longer reports single-digit0
. -
Fixed #5307, where CSS value lists were wrapped in a way that did not preserve semantic structure.
Biome now ensures that CSS value lists follow a more readable format, aligning with Prettier’s behavior.
Before:
* {box-shadow:0 0 0 1px #fff,0 0 0 3.2px rgba(89, 89, 235, 0.25),0 0 0 3.2px rgba(89, 89, 235, 0.25),0 0 0 3.2px red,0 0 0 3.2px rgba(89, 89, 235, 0.25);}After:
* {box-shadow:0 0 0 1px #fff,0 0 0 3.2px rgba(89, 89, 235, 0.25),0 0 0 3.2px rgba(89, 89, 235, 0.25),0 0 0 3.2px red,0 0 0 3.2px rgba(89, 89, 235, 0.25);} -
tsconfig.*.json
files will now be treated the same astsconfig.json
files. -
The
summary
reporter now prints the files processed and the files fixed when passing the--verbose
flag. -
Fixed #5693:
useRegexLiterals
now correctly handle useless escaped character in string literals. -
useRegexLiterals now suggests a correct fix when the pattern contains an escaped anti-slash
\/
.Previously the rule suggested the following fix that led to a syntax error:
new RegExp("\/");/\\//The rule now suggests a correct fix:
new RegExp("\/");/\//Fixed #5487.
-
Fixed
useConsistentCurlyBraces breaks react/no-unescaped-entities rule
Added a check for forbidden characters:
>
,"
,'
and}
. If any of these characters are detected, curly braces will be preserved.Example:
function MyComponent() {return <Foo>Jupiter {">"} Venus</Foo>;} -
The rule
useNamingConvention
now suggests a rename that preserves uppercase if possible.For instance, Biome suggested renaming
HTMLWrapper
ashtmlWrapper
:import HTMLWrapper from "HTMLWrapper.tsx";import htmlWrapper from "HTMLWrapper.tsx";function component() {return <HTMLWrapper> </HTMLWrapper>;return <htmlWrapper> </HTMLWrapper>;}Since both
PascalCase
andCamelCase
are accepted, Biome now suggests renamingHTMLWrapper
asHtmlWrapper
:import HTMLWrapper from "HTMLWrapper.tsx";import HtmlWrapper from "HTMLWrapper.tsx";function component() {return <HTMLWrapper> </HTMLWrapper>;return <HtmlWrapper> </HTMLWrapper>;} -
Fix a parsing error when a
JsxElementName
isJsxMemberExpression
, and aJsLogicalExpreesion
before it without a semicolon.The following case will now not throw error:
import React from "react";let b = 0;function A() {const a = b > 0 && b < 1;return <React.Fragment>{a}</React.Fragment>;} -
Fixed Biome being unable to parse
insert_final_newline = unset
in EditorConfig files. -
Fixed #4530: useArrowFunction now preserves directives.
Previously the rule removed the directives when a function expression was turned into an arrow function. The rule now correctly keeps the directives.
const withDirective = function () {const withDirective = () => {"use server";return 0;} -
Fixed #4855: useSortedClasses now suggests code fixes that match the JSX quote style of the formatter.