Skip to content
Biome
Docs Enterprise Playground
Blog

All versions since 2.0.0
All versions

2.0.0

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 the biome lsp-proxy and biome 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 and noMisleadingCharacterClass checked both regular expression literals like /regex/ and dynamically built regular expressions like new 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 and noMisleadingCharacterClass 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. Run biome migrate to migrate the configuration automatically.

  • Removed include and ignore fields in favor of the new field includes.

    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 and ignore fields. In Biome 2.0, include and ignore are removed and replaced by includes. You can run biome migrate to convert include and ignore into includes 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 replaces ignore 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 with include and ignore, because ignore has priority over include.

    The semantics of * and **/* have changed too. Before, with include and ignore, the glob * was interpreted as **/*. Now, with includes, 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:

    1. Ignore all files ending with .test.js, except special.test.ts.
    2. Ignore all files of the test directory. The test directory is located at the root of the project.
    3. Execute the linter on files in the src directory, that don’t end with .gen.js. The src directory is located at the root of the project.
    4. 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:

    1. There is no way to ignore files and unignore one of them. Thus, we ignore all files ending with .test.js, including special.test.ts.
    2. The configuration ignores all directories named test, including src/test.
    3. 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" } } }
        }
      ]
    }
    1. All files named special.test.ts are unignored because the pattern appear after the pattern that ignore files ending with .test.js.
    2. Only the test directory at the project’s root is ignored because the pattern doesn’t start with **/.
    3. The linter is executed on the src directory at the project’s root only.

    Because includes pattern have a different pattern format than include and ignore we made some adjustments:

    • We added the pattern ** in files.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.

  • 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 to true by default.

    This option is now set to false by default. To get the previous behavior, you have to set checkTypes to true:

    {
      "linter": {
        "rules": {
          "correctness": {
            "noUndeclaredVariables": {
              "level": "on",
              "options": { "checkTypes": true }
            }
          }
        }
      }
    }

  • The rule noUnusedVariables no longer reports unused function parameters. Use noUnusedFunctionParameters.

  • Fixed #5564. noTypeOnlyImportAttributes now ignores files ending with the extension .ts when the type field of package.json is set to commonjs.

  • The Biome formatter no longer adds a trailing comma in .json files, even when json.formatter.trailingCommas is set to true.

  • 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 accept Single and Double anymore.
    • The option --quote-properties doesn’t accept AsNeeded and Preserve anymore.
    • The option --semicolons doesn’t accept AsNeeded and Always anymore.
    • The option --arrow-parenthesis doesn’t accept AsNeeded and Always anymore.
    • The option --trailing-commas doesn’t accept ES5, All and None anymore.
    • The option --attribute-position doesn’t accept Single and Multiline anymore.

  • Removed the option enumMemberCase from the lint rule useNamingConvention.

    enumMemberCase is an option that allows to customize the enforced case for TypeScript’s enum members. The option was introduced prior to the conventions 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 turn enumMemberCase into conventions 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.

  • Removed ROME_BINARY. Use BIOME_BINARY instead.

  • Removed support for legacy suppressions.

    Biome used to support “legacy suppressions” that looked like this:

    // biome-ignore lint(complexity/useWhile): reason

    This 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 or biome.jsonc instead.

  • Removed the option all from the linter.

    The options linter.rules.all and linter.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 option trailingCommas instead:

    {
      "javascript": {
        "formatter": {
          "trailingComma": "es5"
          "trailingCommas": "es5"
        }
      }
    }
    biome format --trailing-comma=es5
    biome 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-unsafe
    biome check --write --unsafe
    biome check --apply
    biome check --write

  • Removed support for assert syntax.

    Biome now longer supports the assert syntax, use the new with syntax instead

    import {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 the rel="noopener" in addition to rel="noreferrer". In addition, an option has been added that allows rel="noreferrer" to be disabled.

    The rule has been moved from the a11y group to the security group.

  • The rule useImportRestrictions has been renamed to noPrivateImports, 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:

    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:

    The following GraphQL rules have been promoted:

    The following JavaScript rules have been promoted:

    Moreover, the following JavaScript rules are now recommended:

    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, or info. Previously, all recommended lint rules had a default severity level set to error. All other lint rules had a default severity level set to warn.

    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.

  • 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 and useExhaustiveDependencies are only recommended when the react 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 the suspicious group

  • The rule useWhile now belongs to the complexity group.

  • The rule useImportExtensions has been updated to suggest actual file extensions instead of guesses based on heuristics.

    As part of this, the suggestedExtensions option has been removed. A simpler, new option called forceJsExtensions has been introduced for those who use tsc’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 to false:

    {
        "linter": {
            "rules": {
                "style": {
                    "useFilenamingConvention": {
                        "level": "on",
                        "options": {
                            "requireAscii": false
                        }
                    }
                    "useFilenamingConvention": {
                        "level": "on",
                        "options": {
                            "requireAscii": false
                        }
                    }
                }
            }
        }
    }

  • Renamed the rule noUnnecessaryContinue to noUselessContinue. Run the command biome migrate to update your configuration.

  • Renamed the rule noMultipleSpacesInRegularExpressionLiterals to noAdjacentSpacesInRegex. Run the command biome migrate to update your configuration.

Minor Changes

  • An option called allowNoReferrer has been added to the noBlankTarget rule.

    By default, noBlankTarget accepts both rel="noopener" and rel="noreferrer" with links that have target="_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 set allowNoReferrer: false to only accept rel="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 to false, 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
    // from
    const 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 "";
      },
    };
    // to
    const 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 with collapse.

  • 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.

    // invalid
    describe.only.each([["a"], ["b"]])("%s", (a) => {});
    it.only.each([["a"], ["b"]])("%s", (a) => {});
    test.only.each([["a"], ["b"]])("%s", (a) => {});
    

    // valid
    describe.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 in settings.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 than off.

  • 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-await
    

      ℹ Rules that can be migrated to an inspired rule using --include-inspired:
    

      - @typescript-eslint/parameter-properties
    

      ℹ Rules that can be migrated to a nursery rule using --include-nursery:
    

      - @typescript-eslint/switch-exhaustiveness-check
    

      ℹ Stylistic rules that the formatter may support (manual migration required):
    

      - semi
    

      ℹ Unsupported rules:
    

      - block-scoped-var
    

    configuration ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    

      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: bug
    import type { MyType } from "my-esm-pkg" with { "resolution-mode": "import" };

    Biome now requires all biome-ignore-start suppressions to have an equivalent biome-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 comment
    import {} from "mod";

  • Added a new propertyAssignment option to the noParameterAssign rule. This option allows to configure whether property assignments on function parameters are permitted. By default, propertyAssignment is set to allow. Setting it to deny 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 another typeof 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 is error, so you would use "on", and then linting a code that uses var, 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 to noVar.

    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 folder root/packages/foo/, and it contains the pattern dist/, only files and directories inside root/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 shorthand
      foo,
    

      // Method shorthand
      method() {
        return "method";
      },
    };

    Alternatively, the rule can be configured to enforce an explicit style:

    const invalidExplicit = {
      // Basic property shorthand violations
      foo: foo,
    

      // Method shorthand violations
      method: 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 of stdout.

    In addition, the --log-level parameter now also accepts a tracing 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.json
    cat tracing.json | jq '. | select(.span.name == "update_module_graph") | { path: .span.path, time_busy: .["time.busy"], time_idle: .["time.idle"] }' > filtered.json

    Now 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 called domains.

    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 enabled
          react: "recommended", // only the recommended rules from this domain are enabled
          solid: "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:

    1. information
    2. warning
    3. 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 on Array)

    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-compiler

    The 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 and lint/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-ignore
    

        2 │ 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 default maxSize:

    {
      "overrides": [{
        "includes": ["dist/**"]
        "files": {
          "maxSize": 20000
        }
      }]
    }

  • Added the new CLI option called --threads to the ci 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=1
    BIOME_THREADS=1 biome ci

  • Added the new rule useObjectSpread, which prefers object spread syntax over Object.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 Biome
    debugger;
     // biome-ignore lint/suspicious/noDoubleEquals: Migration to Biome
    foo == bar;

  • Add an ignoreRestSiblings option into noUnusedVariables.

    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 rule
    console.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: a
    

        3 │ 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 and useImportType 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’s no-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 the with { 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 now camelCase.

  • 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

  • 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 and biome.jsonc, and it will use biome.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 named eslint-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 to Object.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:

    <Suspense
      fallback={
        <>
          <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 named eslint-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 when cssModules 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 with null 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 suppression 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 checks 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 and textDocument/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, and ci.

  • 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 for container-name.

  • Fixed #4841: Shebang and top leading comments in .cjs files are now handled correctly

    Example: shebang only (keep it as is)

    #!/usr/bin/env node

    Example: comments only (keep it as is)

    // comment

    Example: 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";
     // comment
    let some_variable = "some value";

    Example: with shebang and comment

     #!/usr/bin/env node"use strict";
     // comment
     #!/usr/bin/env node
     "use strict";
     // comment
    let 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 works
    const foo0 = function (bar: string) {
      // biome-ignore lint/style/noParameterAssign: even if there are other suppressions inside
      bar = "baz";
    };

  • Add @vitest/eslint-plugin to list of Biome rule sources

  • Fixed useHookAtTopLevel rule to properly detect React components wrapped in memo and forwardRef, 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 inside switch and if 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 diagnostic
              case "number":
                return <div>{x.value}</div>; // no key, emits diagnostic
              default:
                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 nested files 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 and useFilenamingConvention 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 doesn’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 the tech 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 to noDuplicateFields. Run the command biome 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 for role="searchbox" instead of role="search".

  • Support setting indent_size to tab 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:

    1. 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 role

      This change ensures the rule is more accurate and helpful.

    2. 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} />;

    3. 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 is always, it becomes:

    const foo = 3;
    

    [1, 2, 3].map((x) => x * 2);

    when javascript.formatter.semicolons is asNeeded, 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 reported
    new Array();
    // This was not reported
    const xs = new Array();

  • Improved error handling for the container at-rule.

  • Fixed #4665: the LSP previously identified .cjs files as ESM files, making rules like noRedundantUseStrict 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-digit 0.

  • 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 as tsconfig.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 as htmlWrapper:

    import HTMLWrapper from "HTMLWrapper.tsx";
    import htmlWrapper from "HTMLWrapper.tsx";
    

    function component() {
      return <HTMLWrapper> </HTMLWrapper>;
      return <htmlWrapper> </HTMLWrapper>;
    }

    Since both PascalCase and CamelCase are accepted, Biome now suggests renaming HTMLWrapper as HtmlWrapper:

    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 is JsxMemberExpression, and a JsLogicalExpression 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.

2.0.1

Patch Changes

  • #6425 00e97ad Thanks @siketyan! - Fixed #6391: the rule noUselessFragments no longer reports a fragment that contains whitespaces which aren’t trimmed by the runtime.

  • #6417 dd88565 Thanks @ematipico! - Fixed #6360: The following pseudo classes and elements are no longer reported by noUnknownPseudoClass or noUnknownPseudoElement rules.

    • :open
    • ::details-content
    • ::prefix
    • ::search-text
    • ::suffix

  • #6417 dd88565 Thanks @ematipico! - Fixed #6357, where the boolean values weren’t correctly merged when using the extends functionality. Now Biome correctly merges the values.

  • #6417 dd88565 Thanks @ematipico! - Fixed #6341: Fixed an issue where Biome would throw an error for the language tags nb and nn.

  • #6385 94142dd Thanks @siketyan! - Fixed #6377: The rule noSelfCompare now correctly compares two function calls with different arguments.

  • #6417 dd88565 Thanks @ematipico! - Fixed #6278: useExhaustiveDependencies no longer adds duplicated dependencies into the list.

  • #6417 dd88565 Thanks @ematipico! - Fix #6396, where vi.useFakeTimers() and vi.useRealTimers() incorrectly triggered React Hooks-related rules

  • #6417 dd88565 Thanks @ematipico! - Fixed a bug where Biome didn’t correctly discover nested configuration files when using the lint command and the linter is disabled in the root configuration.

  • #6422 594ec50 Thanks @ematipico! - Removed the experimental rename feature from Biome LSP, which caused some issues inside existing editors such as Zed.

  • #6388 c6942d2 Thanks @siketyan! - Fixed #6375: the formatter no longer inserts an extra empty line before a semicolon when it has leading comments.

2.0.2

Patch Changes

  • #6436 ec7c63d Thanks @ematipico! - Fixed an issue where binaries weren’t copied anymore inside the @biomejs/cli-* packages.

2.0.3

Patch Changes

  • #6439 7e4da4e Thanks @ematipico! - Fixed an issue where the correct rights aren’t added to the binary during publishing

  • #6297 cc4b8c9 Thanks @vladimir-ivanov! - Added a new lint useReadonlyClassProperties rule. This rule is a port of ESLint’s prefer-readonly rule.

    Example:

    class Example {
      // All properties below can be marked as readonly
      public constantValue = 42;
      protected initializedInConstructor: string;
      private privateField = true;
    

      constructor(initializedInConstructor: string) {
        this.initializedInConstructor = initializedInConstructor;
      }
    }

2.0.4

Patch Changes

2.0.5

Patch Changes

  • #6461 38862e6 Thanks @ematipico! - Fixed #6419, a regression where stdin mode would create a temporary new file instead of using the one provided by the user. This was an intended regression.

    Now Biome will use the file path passed via --std-file-path, and apply the configuration that matches it.

  • #6480 050047f Thanks @Conaclos! - Fixed #6371. useNamingConvention now checks the string case of objects’ property shorthand.

  • #6477 b98379d Thanks @ematipico! - Fixed an issue where Biome formatter didn’t format consistently CSS value separated by commas.

    .font-heading {
     font-feature-settings: var(--heading-salt), var(--heading-ss06),
       var(--heading-ss11), var(--heading-cv09), var(--heading-liga),
       var(--heading-calt);
    

      font-feature-settings:
        var(--heading-salt), var(--heading-ss06), var(--heading-ss11),
        var(--heading-cv09), var(--heading-liga), var(--heading-calt);
    }

  • #6248 ec7126c Thanks @fireairforce! - Fixed grit pattern matching for different kinds of import statements.

    The grit pattern import $imports from "foo" will match the following code:

    import bar from "foo";
    import { bar } from "foo";
    import { bar, baz } from "foo";

2.0.6

Patch Changes

  • #6557 fd68458 Thanks @ematipico! - Fixed a bug where Biome didn’t provide all the available code actions when requested by the editor.

  • #6511 72623fa Thanks @Conaclos! - Fixed #6492. The organizeImports assist action no longer duplicates a comment at the start of the file when :BLANK_LINE: precedes the first import group.

  • #6557 fd68458 Thanks @ematipico! - Fixed #6287 where Biome Language Server didn’t adhere to the settings.requireConfiguration option when pulling diagnostics and code actions. Note that for this configuration be correctly applied, your editor must support dynamic registration capabilities.

  • #6551 0b63b1d Thanks @Conaclos! - Fixed #6536. useSortedKeys no longer panics in some edge cases where object spreads are involved.

  • #6503 9a8fe0f Thanks @ematipico! - Fixed #6482 where nursery rules that belonged to a domain were incorrectly enabled.

  • #6565 e85761c Thanks @daivinhtran! - Fixed #4677: Now the noUnusedImports rule won’t produce diagnostics for types used in JSDoc comment of exports.

  • #6166 b8cbd83 Thanks @mehm8128! - Added the nursery rule noExcessiveLinesPerFunction. This rule restrict a maximum number of lines of code in a function body.

    The following code is now reported as invalid when the limit of maximum lines is set to 2:

    function foo() {
      const x = 0;
      const y = 1;
      const z = 2;
    }

    The following code is now reported as valid when the limit of maximum lines is set to 3:

    const bar = () => {
      const x = 0;
      const z = 2;
    };

  • #6553 5f42630 Thanks @denbezrukov! - Fixed #6547. Now the Biome CSS parser correctly parses @starting-style when it’s used inside other at-rules. The following example doesn’t raise an error anymore:

    @layer my-demo-layer {
      @starting-style {
        div.showing {
          background-color: red;
        }
      }
    }

  • #6458 05402e3 Thanks @ematipico! - Fixed an issue where the rule useSemanticElements used the incorrect range when positioning suppression comments.

  • #6560 6d8a6b9 Thanks @siketyan! - Fixed #6559: the error message on detected a large file was outdated and referred a removed configuration option files.ignore.

  • #6458 05402e3 Thanks @ematipico! - Fixed #6384. The rule useAltText now emits a diagnostic with a correct range, so suppression comments can work correctly.

  • #6518 7a56288 Thanks @wojtekmaj! - Fixed #6508, where the rule noUselessFragments incorrectly flagged Fragments containing HTML entities as unnecessary.

  • #6517 c5217cf Thanks @arendjr! - Fixed #6515. When using the extends field to extend a configuration from an NPM package, we now accept the condition names "biome" and "default" for exporting the configuration in the package.json.

    This means that where previously your package.json had to contain an export declaration similar to this:

    {
      "exports": {
        ".": "./biome.json"
      }
    }

    You may now use one of these as well:

    {
      "exports": {
        ".": {
          "biome": "./biome.json"
        }
      }
    }

    Or:

    {
      "exports": {
        ".": {
          "default": "./biome.json"
        }
      }
    }

  • #6219 a3a3715 Thanks @huangtiandi1999! - Added new nursery rule noUnassignedVariables, which disallows let or var variables that are read but never assigned.

    The following code is now reported as invalid:

    let x;
    if (x) {
      console.log(1);
    }

    The following code is now reported as valid:

    let x = 1;
    if (x) {
      console.log(1);
    }

  • #6395 f62e748 Thanks @mdevils! - Added the new nursery rule noImplicitCoercion, which disallows shorthand type conversions in favor of explicit type conversion functions.

    Example (Invalid): Boolean conversion using double negation:

    !!foo;
    !!(foo + bar);

    Example (Invalid): Number conversion using unary operators:

    +foo;
    -(-foo);
    foo - 0;
    foo * 1;
    foo / 1;

    Example (Invalid): String conversion using concatenation:

    "" + foo;
    foo + "";
    `` + foo;
    foo += "";

    Example (Invalid): Index checking using bitwise NOT:

    ~foo.indexOf(1);
    ~foo.bar.indexOf(2);

    Example (Valid): Using explicit type conversion functions:

    Boolean(foo);
    Number(foo);
    String(foo);
    foo.indexOf(1) !== -1;

  • #6544 f28b075 Thanks @daivinhtran! - Fixed #6536. Now the rule noUselessFragments produces diagnostics for a top-level useless fragment that is in a return statement.

  • #6320 5705f1a Thanks @mdevils! - Added the new nursery rule useUnifiedTypeSignature, which disallows overload signatures that can be unified into a single signature.

    Overload signatures that can be merged into a single signature are redundant and should be avoided. This rule helps simplify function signatures by combining overloads by making parameters optional and/or using type unions.

    Example (Invalid): Overload signatures that can be unified:

    function f(a: number): void;
    function f(a: string): void;
    interface I {
      a(): void;
      a(x: number): void;
    }

    Example (Valid): Unified signatures:

    function f(a: number | string): void {}
    interface I {
      a(x?: number): void;
    }

    Example (Valid): Different return types cannot be merged:

    interface I {
      f(): void;
      f(x: number): number;
    }

  • #6545 2782175 Thanks @ematipico! - Fixed #6529, where the Biome Language Server would emit an error when the user would open a file that isn’t part of its workspace ( node_modules or external files). Now the language server doesn’t emit any errors and it exits gracefully.

  • #6524 a27b825 Thanks @vladimir-ivanov! - Fixed #6500: The useReadonlyClassProperties rule now correctly marks class properties as readonly when they are assigned in a constructor, setter or method, even if the assignment occurs inside an if or else block.

    The following code is now correctly detected by the rule:

    class Price {
      #price: string;
    

      @Input()
      set some(value: string | number) {
        if (
          value === undefined ||
          value === null ||
          value === "undefined" ||
          value === "null" ||
          Number.isNaN(value)
        ) {
          this.#price = "";
        } else {
          this.#price = "" + value;
        }
      }
    }

  • #6355 e128ea9 Thanks @anthonyshew! - Added a new nursery rule noAlert that disallows the use of alert, confirm and prompt.

    The following code is deemed incorrect:

    alert("here!");

  • #6548 37e9799 Thanks @ematipico! - Fixed #6459, where the Biome LSP was not taking into account the correct settings when applying source.fixAll.biome code action.

2.1.0

Minor Changes

  • #6512 0c0bf82 Thanks @arendjr! - The rule noFloatingPromises can now detect floating arrays of Promises.

    Invalid examples

    // This gets flagged because the Promises are not handled.
    [1, 2, 3].map(async (x) => x + 1);

    Valid examples

    await Promise.all([1, 2, 3].map(async (x) => x + 1));

  • #6637 6918085 Thanks @arendjr! - Type inference is now able to handle the sequence operator ( ,), as well as post- and pre-update operators: ++.

    Example

    let x = 5;
    

    // We now infer that `x++` resolves to a number, while the expression as a whole
    // becomes a Promise:
    (x++, new Promise((resolve) => resolve("comma")));

  • #6752 c9eaca4 Thanks @arendjr! - Fixed #6646: .gitignore files are now picked up even when running Biome from a nested directory, or when the ignore file itself is ignored through files.includes.

  • #6746 90aeead Thanks @arendjr! - biome migrate no longer enables style rules that were recommended in v1, because that would be undesirable for users upgrading from 2.0.

    Users who are upgrading from Biome 1.x are therefore advised to first upgrade to Biome 2.0, and run the migration, before continuing to Biome 2.1 or later.

  • #6583 d415a3f Thanks @arendjr! - Added the nursery rule noMisusedPromises.

    It signals Promises in places where conditionals or iterables are expected.

    Invalid examples

    const promise = Promise.resolve("value");
    

    // Using a `Promise` as conditional is always truthy:
    if (promise) {
      /* ... */
    }
    

    // Spreading a `Promise` has no effect:
    console.log({ foo: 42, ...promise });
    

    // This does not `await` the `Promise`s from the callbacks,
    // so it does not behave as you may expect:
    [1, 2, 3].forEach(async (value) => {
      await fetch(`/${value}`);
    });

    Valid examples

    const promise = Promise.resolve("value");
    

    if (await promise) {
      /* ... */
    }
    

    console.log({ foo: 42, ...(await promise) });

  • #6405 cd4a9bb Thanks @vladimir-ivanov! - Added the ignoreRestSiblings option to the noUnusedFunctionParameters rule.

    This option is used to ignore unused function parameters that are siblings of the rest parameter.

    The default is false, which means that unused function parameters that are siblings of the rest parameter will be reported.

    Example

    {
      "rules": {
        "noUnusedFunctionParameters": ["error", { "ignoreRestSiblings": true }]
      }
    }

  • #6614 0840021 Thanks @arendjr! - We have implemented a more targeted version of the scanner, which ensures that if you provide file paths to handle on the CLI, the scanner will exclude directories that are not relevant to those paths.

    Note that for many commands, such as biome check and biome format, the file paths to handle are implicitly set to the current working directory if you do not provide any path explicitly. The targeted scanner also works with such implicit paths, which means that if you run Biome from a subfolder, other folders that are part of the project are automatically exempted.

    Use cases where you invoke Biome from the root of the project without providing a path, as well as those where project rules are enabled, are not expected to see performance benefits from this.

    Implemented #6234, and fixed #6483 and #6563.

  • #6488 c5ee385 Thanks @ianzone! - nx.json and project.json have been added to the list of well-known files.

  • #6720 52e36ae Thanks @minht11! - Added $ symbol to organizeImports :ALIAS: group.

    import { action } from '$lib' will be treated as alias import.

Patch Changes

  • #6712 2649ac6 Thanks @sterliakov! - Fixed #6595: Biome now supports // biome-ignore-all file-level suppressions in files that start with a shebang (#!).

  • #6758 28dc49e Thanks @arendjr! - Fixed #6573: Grit plugins can now match bare imports.

    Example

    The following snippet:

    `import $source`

    will now match:

    import "main.css";

  • #6550 b424f46 Thanks @arendjr! - Type inference is now able to handle logical expressions: &&, ||, and ??.

    Examples

    // We can now infer that because `true` is truthy, the entire expression
    // evaluates to a `Promise`.
    true && Promise.reject("logical operator bypass");
    

    // And we know that this doesn't:
    false && Promise.reject("logical operator bypass");
    

    // Truthiness, falsiness, and non-nullishness can all be determined on more
    // complex expressions as well. So the following also works:
    type Nullish = null | undefined;
    

    type Params = {
      booleanOption: boolean | Nullish;
      falsyOption: false | Nullish;
    };
    

    function foo({ booleanOption, falsyOption }: Params) {
      // This may be a Promise:
      booleanOption ?? Promise.reject("logical operator bypass");
    

      // But this never is:
      falsyOption && Promise.reject("logical operator bypass");
    }

  • #6413 4aa0e50 Thanks @wojtekmaj! - Improved error message in useDateNow rule.

  • #6673 341e062 Thanks @dyc3! - Fixed a case where the HTML formatter would mangle embedded language tags if whitespaceSensitivity was set to strict

  • #6642 a991229 Thanks @unvalley! - Fixed #4494: The noSecrets rule now correctly uses the entropyThreshold option to detect secret like strings.

  • #6520 0c43545 Thanks @arendjr! - Type inference is now able to handle ternary conditions in type aliases.

    Note that we don’t attempt to evaluate the condition itself. The resulting type is simply a union of both conditional outcomes.

    Example

    type MaybeResult<T> = T extends Function ? Promise<string> : undefined;
    

    // We can now detect this function _might_ return a `Promise`:
    function doStuff<T>(input: T): MaybeResult<T> {
      /* ... */
    }

  • #6711 1937691 Thanks @sterliakov! - Fixed #6654: Fixed range highlighting of <explanation> placeholder in inline suppression block comments.

  • #6756 d12b26f Thanks @dyc3! - Fixed #6669: Added an exception to noUnusedImports to allow type augmentation imports.

    import type {} from "@mui/lab/themeAugmentation";

  • #6643 df15ad6 Thanks @skewb1k!

  • Fixed #4994: LSP server registered some capabilities even when the client did not support dynamic registration.

  • #6599 5e611fa Thanks @vladimir-ivanov! - Fixed #6380: The noFocusedTests rule now correctly displays the function name in the diagnostic message when a test is focused.

    Every instance of a focused test function (like fdescribe, fit, ftest and only) had the word ‘only’ hardcoded. This has been updated to use the actual function name, so the message is now more accurate and specific.

    Example for fdescribe:

      i The 'fdescribe' method is often used for debugging or during implementation.
    

      i Consider removing 'f' prefix from 'fdescribe' to ensure all tests are executed.

  • #6671 0c9ab43 Thanks @vladimir-ivanov! - Fixed #6634: The useReadonlyClassProperties rule now correctly flags mutations in class getters and in arrow functions within class properties.

    Examples:

    class GetterWithMutationValue {
      #value: string;
    

      get value() {
        if (!this.#value) {
          this.#value = "defaultValue";
        }
    

        return this.#value;
      }
    }
    class ClassPropertyArrowFunctionWithMutation {
      private bar: string | null = null;
    

      readonly action = () => {
        this.bar = "init";
      };
    }

  • #6682 ca04cea Thanks @ematipico! - Fixed #6668: Biome Assist is now enabled by default for CSS files.

  • #6525 66b089c Thanks @arendjr! - Type inference can now infer the return types of functions and methods without annotations.

    Examples

    const sneakyObject = {
      doSomething() {
        return Promise.resolve("This is a floating promise!");
      },
    };
    

    // We can now detect that `doSomething()` returns a `Promise`.
    sneakyObject.doSomething();

  • #6531 c06df79 Thanks @arendjr! - Biome’s type inference now detects the type of properties with getters.

    Examples

    const sneakyObject2 = {
      get something() {
        return new Promise((_, reject) => reject("This is a floating promise!"));
      },
    };
    // We now detect this is a Promise:
    sneakyObject2.something;

  • #6587 a330fcc Thanks @Conaclos! - organizeImports is now able to sort named specifiers and import attributes with bogus nodes.

  • #6618 6174869 Thanks @Shinyaigeek! - Fixed #6610: JSON import attributes are now correctly detected when they contain extra whitespace.

  • #6753 fce5d2c Thanks @dyc3! - Improved the error messages when Biome is provided incompatible arguments on the CLI.

  • #6587 a330fcc Thanks @Conaclos! - Fixed #6491: The action of useSortedKeys removed comments or wrongly transferred them to distinct nodes.

  • #6696 92964a7 Thanks @unvalley! - Fixed #6633: The noImplicitCoercion rule no longer reports diagnostics for 1 / value expressions.

    1 / value; // no error

  • #6683 43d871e Thanks @ematipico! - Fixed #6537: Biome no longer removes the trailing comma from JSON files when formatter.json.trailingCommas is explicitly set to "all".

  • #6693 bfdce0b Thanks @dyc3! - Fixed #6691: The HTML parser will now consider . to be a valid character for tag names.

  • #6716 ead03d1 Thanks @siketyan! - The Biome LSP server no longer responds with an error for a textDocument/codeActions request when Biome doesn’t support a feature for the file (e.g. Code actions aren’t supported in GritQL files).

  • #6679 7bf9a60 Thanks @marko-hologram! - Fixed #6638: JavaScript formatter overrides options now correctly override expand option. JSON formatter overrides options now correctly override bracketSpacing and expand options.

  • #6717 7f5b541 Thanks @siketyan! - Fixed #6688: the noUselessFragments no longer reports <Fragment /> elements that includes HTML character entities.

  • #6600 853e1b5 Thanks @daivinhtran! - Fixed #4677: The noUnusedImports rule won’t produce diagnostics for types used in comments of static members anymore.

  • #6662 3afc804 Thanks @arendjr! - If a nested configuration file is ignored by the root configuration, it will now actually be ignored.

    Biome has an exception in place for configuration files so they cannot be ignored, because the configuration files are vital to Biome itself. But this exception was incorrectly applied to nested configurations as well. Now only the root configuration is exempt from being ignored.

  • #6596 c0718ca Thanks @ematipico! - Fixed #6566: Biome no longer errors when using the option --files-ignore-unknown=true in stdin mode.

    Biome has also become less strict when using --stdin-file-path in stdin mode. It will no longer error if the file path doesn’t contain an extension, but instead it will return the original content.

  • #6562 153eda7 Thanks @vladimir-ivanov! - Added the nursery rule noMagicNumbers. The rule detects and reports the use of “magic numbers” — numeric literals that are used directly in code without being assigned to a named constant.

    Example

    let total = price * 1.23; // Magic number for tax rate will highlight 1.23 as magic number

  • #6663 af78d6d Thanks @ematipico! - Fixed #6656: Biome now correctly formats HTML void elements such as <meta> when they contain a self-closing slash.

    <meta foo="bar" />
    <meta foo="bar">

  • #6732 31e4396 Thanks @vladimir-ivanov! - Resolved #6281: Improved performance of handling package.json files in the scanner.

  • #6625 19cb475 Thanks @arendjr! - Fixed #6616: Fixed an issue with extending configurations that contained an explicit root field while the configuration in the project did not.

  • #6650 19aab18 Thanks @sterliakov! - Fixed #6621: Improved handling of multiple adjacent line suppressions. Biome now handles such suppressions separately, tracking whether each one is used.

  • #6700 cdd6e17 Thanks @denbezrukov! - Fixed #6680: Biome incorrectly formatted container-style queries by inserting misplaced spaces.

    @container style (--responsive: true) {}
    @container style(--responsive: true) {}

  • #6709 ecf3954 Thanks @dyc3! - Fixed #6038: Fixed a false positive in noShadow where a function parameter in a type definition was erroneously flagged as a violation.

  • #6593 a4acbb7 Thanks @arendjr! - Type inference is now able to handle ternary conditions in expressions.

    Examples

    const condition = Math.random() > -1; // Always true, but dynamic to linter
    

    // We now detect that this may return a `Promise`.
    condition ? Promise.reject("ternary bypass") : null;
    

    // On the other hand, we know the following is never a `Promise`:
    const alwaysFalsy = 0;
    alwaysFalsy ? Promise.reject("ternary bypass") : null;

  • #6428 4b501d3 Thanks @siketyan! - Added MemoryFileSystem to the WASM API.

    You can now insert a file from your JS code:

    import { MemoryFileSystem, Workspace } from "@biomejs/wasm-web";
    

    const fs = new MemoryFileSystem();
    const workspace = Workspace.withFileSystem(fs);
    

    fs.insert("/index.js", new TextEncoder().encode("let foo = 1;"));
    fs.remove("/index.js");

  • #6594 626d4a1 Thanks @ematipico! - Fixed #6528: Biome didn’t return the correct output when applying source.fixAll.biome inside Astro/Vue/Svelte files that contained safe fixed.

2.1.1

Patch Changes

  • #6781 9bbd34f Thanks @siketyan! - Fixed the FileFeaturesResult interface in the WASM API was defined as a mapped object but the actual value was a Map object.

  • #6761 cf3c2ce Thanks @dyc3! - Fixed #6759, a false positive for noFocusedTests that was triggered by calling any function with the name fit on any object.

    The following code will now pass the noFocusedTests rule:

    import foo from "foo";
    foo.fit();

2.1.2

Patch Changes

  • #6865 b35bf64 Thanks @denbezrukov! - Fix #6485: Handle multiple semicolons correctly in blocks (#6485)

    div {
      box-sizing: border-box;
      color: red;
    }

  • #6798 3579ffa Thanks @dyc3! - Fixed #6762, Biome now knows that ~/.config/zed/settings.json and ~/.config/Code/User/settings.json allows comments by default.

  • #6839 4cd62d8 Thanks @ematipico! - Fixed #6838, where the Biome File Watcher incorrectly watched and stored ignored files, causing possible memory leaks when those files were dynamically created (e.g. built files).

  • #6879 0059cd9 Thanks @denbezrukov! - Refactor: remove one level of indirection for CSS declarations with semicolon Previously, accessing a declaration from a list required an extra step:

    item
    .as_any_css_declaration_with_semicolon()
    .as_css_declaration_with_semicolon()

    Now, it can be done directly with:

    item.as_css_declaration_with_semicolon()

  • #6839 4cd62d8 Thanks @ematipico! - Fixed a bug where the Biome Language Server didn’t correctly ignore specific files when vcs.useIgnoreFile is set to true.

  • #6884 5ff50f8 Thanks @arendjr! - Improved the performance of noImportCycles by ~30%.

  • #6903 241dd9e Thanks @arendjr! - Fixed #6829: Fixed a false positive reported by useImportExtensions when importing a .js file that had a matching .d.ts file in the same folder.

  • #6846 446112e Thanks @darricheng! - Fixed an issue where biome was using the wrong string quotes when the classes string has quotes, resulting in invalid code after applying the fix.

  • #6823 eebc48e Thanks @arendjr! - Improved #6172: Optimised the way function arguments are stored in Biome’s type inference. This led to about 10% performance improvement in RedisCommander.d.ts and about 2% on @next/font type definitions.

  • #6878 3402976 Thanks @ematipico! - Fixed a bug where the Biome Language Server would apply an unsafe fix when using the code action quickfix.biome.

    Now Biome no longer applies an unsafe code fix when using the code action quickfix.biome.

  • #6794 4d5fc0e Thanks @vladimir-ivanov! - Fixed #6719: The noInvalidUseBeforeDeclaration rule covers additional use cases.

    Examples:

    type Bar = { [BAR]: true };
    const BAR = "bar";
    interface Bar {
      child: { grandChild: { [BAR]: typeof BAR; enumFoo: EnumFoo } };
    }
    const BAR = "bar";
    enum EnumFoo {
      BAR = "bar",
    }

  • #6863 531e97e Thanks @dyc3! - Biome now considers whether the linter is enabled when figuring out how the project should be scanned. Resolves #6815.

  • #6832 bdbc2b1 Thanks @togami2864! - Fixed #6165: Fixed false negative in noUnusedPrivateClassMembers rule when checking member usage in classes

  • #6839 4cd62d8 Thanks @ematipico! - Fixed a bug where the root ignore file wasn’t correctly loaded during the scanning phase, causing false positives and incorrect expectations among users.

    Now, when using vcs.useIgnoreFile, the **the globs specified in the ignore file from the project root ** will have the same semantics as the files.includes setting of the root configuration.

    Refer to the relative web page to understand how they work.

  • #6898 5beb024 Thanks @arendjr! - Fixed #6891: Improved type inference for array indices.

    Example:

    const numbers: number[];
    numbers[42]; // This now infers to `number | undefined`.

  • #6809 8192451 Thanks @arendjr! - Fixed #6796: Fixed a false positive that happened in noFloatingPromises when calling functions that were declared as part of for ... of syntax inside async functions.

    Instead, the variables declared inside for ... of loops are now correctly inferred if the expression being iterated evaluates to an Array (support for other iterables will follow later).

    Invalid example

    const txStatements: Array<(tx) => Promise<any>> = [];
    

    db.transaction((tx: any) => {
      for (const stmt of txStatements) {
        // We correctly flag this resolves to a `Promise`:
        stmt(tx);
      }
    });

    Valid example

    async function valid(db) {
      const txStatements: Array<(tx: any) => void> = [(tx) => tx.insert().run()];
    

      db.transaction((tx: any) => {
        for (const stmt of txStatements) {
          // We don't flag a false positive here anymore:
          stmt(tx);
        }
      });
    }

  • #6757 13a0818 Thanks @mdevils! - Added the rule noVueReservedProps, resolves #6309.

    It prevents the use of reserved Vue prop names such as key and ref which can cause conflicts and unexpected behavior in Vue components.

    Invalid example
    import { defineComponent } from "vue";
    

    export default defineComponent({
      props: ["ref", "key", "foo"],
    });
    <script setup>
    defineProps({
      ref: String,
      key: String,
      foo: String,
    });
    </script>
    Valid examples
    import { defineComponent } from "vue";
    

    export default defineComponent({
      props: ["foo"],
    });
    <script setup>
    defineProps({ foo: String });
    </script>

  • #6840 1a57b51 Thanks @denbezrukov! - Allow multiple identifiers in ::part() pseudo-element selector.

    ::part(first second) {
    }

  • #6845 4fd44ec Thanks @arendjr! - Fixed #6510: The scanner no longer shows diagnostics on inaccessible files unless --verbose is used.

  • #6844 b7e2d4d Thanks @sterliakov! - Fixed #6837: Fixed regression with multiple consecutive line suppression comments using instances (like // biome-ignore lint/correctness/useExhaustiveDependencies(depName): reason).

  • #6818 5f3f5a6 Thanks @siketyan! - Fixed an issue where textDocument/codeAction in the LSP could respond with outdated text edits after the workspace watcher observed outdated changes to the file.

  • #6804 3e6ab16 Thanks @arendjr! - noFloatingPromises will no longer suggest to add await keyword inside synchronous callbacks nested inside async functions.

  • #6901 c9e969a Thanks @arendjr! - Fixed #6777: Fixed type inference handling of this to avoid infinite recursion.

    Thanks to @sterliakov for the thorough investigation!

  • #6855 d1581c7 Thanks @vladimir-ivanov! - Fixed #6775: useReadonlyClassProperties now also captures mutations inside function arguments.

    Example:

    class Counter {
      private counter: number;
      count() {
        console.log(this.counter++);
        const counterString = `${this.counter++}`;
      }
    }

  • #6839 4cd62d8 Thanks @ematipico! - Fixed a bug where Biome didn’t throw any error when vcs.useIgnoreFile is set to true, and there wasn’t any ignore file read. Now Biome correctly throws an error if no ignore files are found.

  • #6911 6d68074 Thanks @arendjr! - Fixed #6838: Reduce resource consumption in the Biome Language Server by using non-recursive filesystem watchers instead of recursive ones.

    Watchers are responsible for notifying Biome of changes to files in the filesystem. We used to set up a single recursive watcher, but that meant that Biome would receive filesystem notifications for all files in your project, even for ignored folders such as build/ or dist/ folders.

    With this patch, we set up non-recursive watchers only for the folders that are relevant to a project.

    Related to this, we also solved an issue where incoming notifications were incorrectly filtered, causing ignored files to be processed and stored in our module graph anyway.

2.1.3

Patch Changes

  • #7057 634a667 Thanks @mdevils! - Added the rule noVueReservedKeys, which prevents the use of reserved Vue keys.

    It prevents the use of Vue reserved keys such as those starting with like $el, $data, $props) and keys starting with \_ in data properties, which can cause conflicts and unexpected behavior in Vue components.

    Invalid example
    <script>
    export default {
      data: {
        $el: "",
        _foo: "bar",
      },
    };
    </script>
    <script>
    export default {
      computed: {
        $data() {
          return this.someData;
        },
      },
    };
    </script>
    Valid examples
    <script>
    export default {
      data() {
        return {
          message: "Hello Vue!",
          count: 0,
        };
      },
    };
    </script>
    <script>
    export default {
      computed: {
        displayMessage() {
          return this.message;
        },
      },
    };
    </script>

  • #6941 734d708 Thanks @JamBalaya56562! - Added @eslint-react/no-nested-component-definitions as a rule source for noNestedComponentDefinitions. Now it will get picked up by biome migrate --eslint.

  • #6463 0a16d54 Thanks @JamBalaya56562! - Fixed a website link for the useComponentExportOnlyModules linter rule to point to the correct URL.

  • #6944 e53f2fe Thanks @sterliakov! - Fixed #6910: Biome now ignores type casts and assertions when evaluating numbers for noMagicNumbers rule.

  • #6991 476cd55 Thanks @denbezrukov! - Fixed #6973: Add support for parsing the :active-view-transition-type() pseudo-class

    :active-view-transition-type(first second) {
    }

  • #6992 0b1e194 Thanks @ematipico! - Added a new JSON rule called noQuickfixBiome, which disallow the use of code action quickfix.biome inside code editor settings.

  • #6943 249306d Thanks @JamBalaya56562! - Fixed @vitest/eslint-plugin source url.

  • #6947 4c7ed0f Thanks @JamBalaya56562! - Fixed ESLint migration for the rule prefer-for from eslint-plugin-solid to Biome’s useForComponent.

  • #6976 72ebadc Thanks @siketyan! - Fixed #6692: The rules noUnusedVariables and noUnusedFunctionParameters no longer cause an infinite loop when the suggested name is not applicable (e.g. the suggested name is already declared in the scope).

  • #6990 333f5d0 Thanks @rvanlaarhoven! - Fixed the documentation URL for lint/correctness/noUnknownPseudoClass

  • #7000 4021165 Thanks @harxki! - Fixed #6795: noUnassignedVariables now correctly recognizes variables used in JSX ref attributes.

  • #7044 b091ddf Thanks @ematipico! - Fixed #6622, now the rule useSemanticElements works for JSX self-closing elements too.

  • #7014 c4864e8 Thanks @siketyan! - Fixed #6516: The biome migrate command no longer break the member list with trailing comments.

  • #6979 29cb6da Thanks @unvalley! - Fixed #6767: useSortedClasses now correctly removes leading and trailing whitespace in className.

    Previously, trailing spaces in className were not fully removed.

    // Think we have this code:
    <div className="text-sm font-bold            " />
    

    // Before: applied fix, but a trailing space was preserved
    <div className="font-bold text-sm " />
    

    // After: applied fix, trailing spaces removed
    <div className="font-bold text-sm" />

  • #7055 ee4828d Thanks @dyc3! - Added the nursery rule useReactFunctionComponents. This rule enforces the preference to use function components instead of class components.

    Valid:

    function Foo() {
      return <div>Hello, world!</div>;
    }

    Invalid:

    class Foo extends React.Component {
      render() {
        return <div>Hello, world!</div>;
      }
    }

  • #6924 2d21be9 Thanks @ematipico! - Fixed #113, where the Biome Language Server didn’t correctly update the diagnostics when the configuration file is modified in the editor. Now the diagnostics are correctly updated every time the configuration file is modified and saved.

  • #6931 e6b2380 Thanks @arendjr! - Fixed #6915: useHookAtTopLevel no longer hangs when rules call themselves recursively.

  • #7012 01c0ab4 Thanks @siketyan! - Fixed #5837: Invalid suppression comments such as biome-ignore-all-start or biome-ignore-all-end no longer causes a panic.

  • #6949 48462f8 Thanks @fireairforce! - Support parse import defer(which is a stage3 proposal). The syntax look like this:

    import defer * as foo from "<specifier>";

  • #6938 5feb5a6 Thanks @vladimir-ivanov! - Fixed #6919 and #6920: useReadonlyClassProperties now does checks for mutations in async class methods.

    Example:

    class Counter3 {
      private counter: number;
      async count() {
        this.counter = 1;
        const counterString = `${this.counter++}`;
      }
    }

  • #6942 cfda528 Thanks @sterliakov! - Fixed #6939. Biome now understands this binding in classes outside of methods.

2.1.4

Patch Changes

  • #7121 b9642ab Thanks @arendjr! - Fixed #7111: Imported symbols using aliases are now correctly recognised.

  • #7103 80515ec Thanks @omasakun! - Fixed #6933 and #6994.

    When the values of private member assignment expressions, increment expressions, etc. are used, those private members are no longer marked as unused.

  • #6887 0cc38f5 Thanks @ptkagori! - Added the noQwikUseVisibleTask rule to Qwik.

    This rule is intended for use in Qwik applications to warn about the use of useVisibleTask$() functions which require careful consideration before use.

    Invalid:

    useVisibleTask$(() => {
      console.log("Component is visible");
    });

    Valid:

    useTask$(() => {
      console.log("Task executed");
    });

  • #7084 50ca155 Thanks @ematipico! - Added the new nursery rule noUnnecessararyConditions, which detects whenever some conditions don’t change during the life cycle of the program, and truthy or false, hence deemed redundant.

    For example, the following snippets will trigger the rule:

    // Always truthy literal conditions
    if (true) {
      console.log("always runs");
    }
    // Unnecessary condition on constrained string type
    function foo(arg: "bar" | "baz") {
      if (arg) {
        // This check is unnecessary
      }
    }

  • #6887 0cc38f5 Thanks @ptkagori! - Added the useImageSize rule to Biome.

    The useImageSize rule enforces the use of width and height attributes on <img> elements for performance reasons. This rule is intended to prevent layout shifts and improve Core Web Vitals by ensuring images have explicit dimensions.

    Invalid:

    <img src="/image.png" />
    <img src="https://example.com/image.png" />
    <img src="/image.png" width="200" />
    <img src="/image.png" height="200" />

    Valid:

    <img width="200" height="600" src="/static/images/portrait-01.webp" />
    <img width="100" height="100" src="https://example.com/image.png" />

  • #6887 0cc38f5 Thanks @ptkagori! - Added the useAnchorHref rule to Biome.

    The useAnchorHref rule enforces the presence of an href attribute on <a> elements in JSX. This rule is intended to ensure that anchor elements are always valid and accessible.

    Invalid:

    <a>Link</a>
    <a target="_blank">External</a>

    Valid:

    <a href="/home">Home</a>
    <a href="https://example.com" target="_blank">
      External
    </a>

  • #7100 29fcb05 Thanks @Jayllyz! - Added the rule noNonNullAssertedOptionalChain.

    This rule prevents the use of non-null assertions (!) immediately after optional chaining expressions ( ?.). Optional chaining is designed to safely handle nullable values by returning undefined when the chain encounters null or undefined. Using a non-null assertion defeats this purpose and can lead to runtime errors.

    // Invalid - non-null assertion after optional chaining
    obj?.prop!;
    obj?.method()!;
    obj?.[key]!;
    obj?.prop!;
    

    // Valid - proper optional chaining usage
    obj?.prop;
    obj?.method();
    obj?.prop ?? defaultValue;
    obj!.prop?.method();

  • #7129 9f4538a Thanks @drwpow! - Removed option, combobox, listbox roles from useSemanticElements suggestions

  • #7106 236deaa Thanks @arendjr! - Fixed #6985: Inference of return types no longer mistakenly picks up return types of nested functions.

  • #7102 d3118c6 Thanks @omasakun! - Fixed #7101: noUnusedPrivateClassMembers now handles members declared as part of constructor arguments:

    1. If a class member defined in a constructor argument is only used within the constructor, it removes the private modifier and makes it a plain method argument.
    2. If it is not used at all, it will prefix it with an underscore, similar to noUnusedFunctionParameter.

  • #7104 5395297 Thanks @harxki! - Reverting to prevent regressions around ref handling

  • #7143 1a6933a Thanks @siketyan! - Fixed #6799: The noImportCycles rule now ignores type-only imports if the new ignoreTypes option is enabled (enabled by default).

    [!WARNING] Breaking Change: The noImportCycles rule no longer detects import cycles that include one or more type-only imports by default. To keep the old behaviour, you can turn off the ignoreTypes option explicitly:

    {
      "linter": {
        "rules": {
          "nursery": {
            "noImportCycles": {
              "options": {
                "ignoreTypes": false
              }
            }
          }
        }
      }
    }

  • #7099 6cc84cb Thanks @arendjr! - Fixed #7062: Biome now correctly considers extended configs when determining the mode for the scanner.

  • #6887 0cc38f5 Thanks @ptkagori! - Added the useQwikClasslist rule to Biome.

    This rule is intended for use in Qwik applications to encourage the use of the built-in class prop (which accepts a string, object, or array) instead of the classnames utility library.

    Invalid:

    <div class={classnames({ active: true, disabled: false })} />

    Valid:

    <div classlist={{ active: true, disabled: false }} />

  • #7019 57c15e6 Thanks @fireairforce! - Added support in the JS parser for import source(a stage3 proposal). The syntax looks like:

    import source foo from "<specifier>";

  • #7053 655049e Thanks @jakeleventhal! - Added the useConsistentTypeDefinitions rule.

    This rule enforces consistent usage of either interface or type for object type definitions in TypeScript.

    The rule accepts an option to specify the preferred style:

    • interface (default): Prefer using interface for object type definitions
    • type: Prefer using type for object type definitions

    Examples:

    // With default option (interface)
    // ❌ Invalid
    type Point = { x: number; y: number };
    

    // ✅ Valid
    interface Point {
      x: number;
      y: number;
    }
    

    // With option { style: "type" }
    // ❌ Invalid
    interface Point {
      x: number;
      y: number;
    }
    

    // ✅ Valid
    type Point = { x: number; y: number };

    The rule will automatically fix simple cases where conversion is straightforward.

2.2.0

Minor Changes

  • #5506 1f8755b Thanks @sakai-ast! - The noRestrictedImports rule has been enhanced with a new patterns option. This option allows for more flexible and powerful import restrictions using gitignore-style patterns.

    You can now define patterns to restrict entire groups of modules. For example, you can disallow imports from any path under import-foo/ except for import-foo/baz.

    {
      "options": {
        "patterns": [
          {
            "group": ["import-foo/*", "!import-foo/baz"],
            "message": "import-foo is deprecated, except for modules in import-foo/baz."
          }
        ]
      }
    }

    Invalid examples

    import foo from "import-foo/foo";
    import bar from "import-foo/bar";

    Valid examples

    import baz from "import-foo/baz";

    Additionally, the patterns option introduces importNamePattern to restrict specific import names using regular expressions. The following example restricts the import names that match x , y or z letters from modules under import-foo/.

    {
      "options": {
        "patterns": [
          {
            "group": ["import-foo/*"],
            "importNamePattern": "[xyz]"
          }
        ]
      }
    }

    Invalid examples

    import { x } from "import-foo/foo";

    Valid examples

    import { foo } from "import-foo/foo";

    Furthermore, you can use the invertImportNamePattern boolean option to reverse this logic. When set to true, only the import names that match the importNamePattern will be allowed. The following configuration only allows the import names that match x , y or z letters from modules under import-foo/.

    {
      "options": {
        "patterns": [
          {
            "group": ["import-foo/*"],
            "importNamePattern": "[xyz]",
            "invertImportNamePattern": true
          }
        ]
      }
    }

    Invalid examples

    import { foo } from "import-foo/foo";

    Valid examples

    import { x } from "import-foo/foo";

  • #6506 90c5d6b Thanks @nazarhussain! - Allow customization of the sort order for different sorting actions. These actions now support a sort option:

    For each of these options, the supported values are the same:

    1. natural. Compares two strings using a natural ASCII order. Uppercase letters come first (e.g. A < a < B < b) and number are compared in a human way (e.g. 9 < 10). This is the default value.
    2. **lexicographic **. Strings are ordered lexicographically by their byte values. This orders Unicode code points based on their positions in the code charts. This is not necessarily the same as “alphabetical” order, which varies by language and locale.

  • #7159 df3afdf Thanks @ematipico! - Added the new rule useBiomeIgnoreFolder. Since v2.2, Biome correctly prevents the indexing and crawling of folders.

    However, the correct pattern has changed. This rule attempts to detect incorrect usage, and promote the new pattern:

    biome.json
    {
      "files": {
        "includes": [
          "!dist/**",
          "!**/fixtures/**",
          "!dist",
          "!**/fixtures",
        ]
      }
    }

  • #6989 85b1128 Thanks @arendjr! - Fixed minor inconsistencies in how files.includes was being handled.

    Previously, Biome sometimes failed to properly ignore the contents of a folder if you didn’t specify the /** at the end of a glob pattern. This was unfortunate, because it meant we still had to traverse the folder and then apply the glob to every entry inside it.

    This is no longer an issue and we now recommend to ignore folders without using the /** suffix.

  • #7118 a78e878 Thanks @avshalomt2! - Added support for .graphqls files. Biome can now format and lint GraphQL files that have the extension .graphqls

  • #6159 f02a296 Thanks @bavalpey! - Added a new option to Biome’s JavaScript formatter, javascript.formatter.operatorLinebreak, to configure whether long lines should be broken before or after binary operators.

    For example, the following configuration:

    {
      formatter: {
        javascript: {
          operatorLinebreak: "before", // defaults to "after"
        },
      },
    }

    Will cause this JavaScript file:

    const VERY_LONG_CONDITION_1234123412341234123412341234 = false;
    

    if (
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234
    ) {
      console.log("DONE");
    }

    to be formatted like this:

    const VERY_LONG_CONDITION_1234123412341234123412341234 = false;
    

    if (
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234 &&
      VERY_LONG_CONDITION_1234123412341234123412341234
    ) {
      console.log("DONE");
    }

  • #7137 a653a0f Thanks @ematipico! - Promoted multiple lint rules from nursery to stable groups and renamed several rules for consistency.

    The following rules have been promoted from nursery to stable groups:

    CSS
    GraphQL
    JavaScript/TypeScript

    Renamed rules

    The following rules have been renamed during promotion. The migration tool will automatically update your configuration:

    Configuration files using the old rule names will need to be updated. Use the migration tool to automatically update your configuration:

    Terminal window
    biome migrate --write

  • #7159 df3afdf Thanks @ematipico! - Added the new rule noBiomeFirstException. This rule prevents the incorrect usage of patterns inside files.includes.

    This rule catches if the first element of the array contains !. This mistake will cause Biome to analyze no files:

    biome.json
    {
      files: {
        includes: ["!dist/**"], // this is an error
      },
    }

  • #6923 0589f08 Thanks @ptkagori! - Added Qwik Domain to Biome

    This release introduces **Qwik domain support ** in Biome, enabling Qwik developers to use Biome as a linter and formatter for their projects.

  • #6989 85b1128 Thanks @arendjr! - Fixed #6965: Implemented smarter scanner for project rules.

    Previously, if project rules were enabled, Biome’s scanner would scan all dependencies regardless of whether they were used by/reachable from source files or not. While this worked for a first version, it was far from optimal.

    The new scanner first scans everything listed under the files.includes setting, and then descends into the dependencies that were discovered there, including transitive dependencies. This has three main advantages:

    • Dependencies that are not reachable from your source files don’t get indexed.
    • Dependencies that have multiple type definitions, such as those with separate definitions for CommonJS and ESM imports, only have the relevant definitions indexed.
    • If vcs.useIgnoreFile is enabled, .gitignore gets respected as well. Assuming you have folders such as build/ or dist/ configured there, those will be automatically ignored by the scanner.

    The change in the scanner also has a more nuanced impact: Previously, if you used files.includes to ignore a file in an included folder, the scanner would still index this file. Now the file is fully ignored, unless you import it.

    As a user you should notice better scanner performance (if you have project rules enabled), and hopefully you need to worry less about configuring files.experimentalScannerIgnores. Eventually our goal is still to deprecate that setting, so if you’re using it today, we encourage you to see which ignores are still necessary there, and whether you can achieve the same effect by ignoring paths using files.includes instead.

    None of these changes affect the scanner if no project rules are enabled.

  • #6731 d6a05b5 Thanks @ematipico! - The --reporter=summary has been greatly enhanced. It now shows the list of files that contains violations, the files shown are clickable and can be opened from the editor.

    Below an example of the new version:

    reporter/parse ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    

      i The following files have parsing errors.
    

      - index.css
    

    reporter/format ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    

      i The following files needs to be formatted.
    

      - index.css
      - index.ts
      - main.ts
    

    reporter/violations ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    

      i Some lint rules or assist actions reported some violations.
    

      Rule Name                                        Diagnostics
    

      lint/correctness/noUnknownFunction               14 (2 error(s), 12 warning(s), 0 info(s))
      lint/suspicious/noImplicitAnyLet                 16 (12 error(s), 4 warning(s), 0 info(s))
      lint/suspicious/noDoubleEquals                   8 (8 error(s), 0 warning(s), 0 info(s))
      assist/source/organizeImports                    2 (2 error(s), 0 warning(s), 0 info(s))
      lint/suspicious/noRedeclare                      12 (12 error(s), 0 warning(s), 0 info(s))
      lint/suspicious/noDebugger                       8 (8 error(s), 0 warning(s), 0 info(s))

  • #6896 527db7f Thanks @ematipico! - Added new functions to the @biomejs/wasm-* packages:

    • fileExists: returns whether the input file exists in the workspace.
    • isPathIgnored: returns whether the input path is ignored.
    • updateModuleGraph: updates the internal module graph of the input path.
    • getModuleGraph: it returns a serialized version of the internal module graph.
    • scanProject: scans the files and directories in the project to build the internal module graph.

  • #6398 d1a315d Thanks @josh-! - Added support for tracking stable results in user-provided React hooks that return objects to useExhaustiveDependencies to compliment existing support for array return values. For example:

    biome.json
    {
      // rule options
      useExhaustiveDependencies: {
        level: "error",
        options: {
          hooks: [
            {
              name: "useCustomHook",
              stableResult: ["setMyState"],
            },
          ],
        },
      },
    }

    This will allow the following to be validated:

    const { myState, setMyState } = useCustomHook();
    const toggleMyState = useCallback(() => {
      setMyState(!myState);
    }, [myState]); // Only `myState` needs to be specified here.

  • #7201 2afaa49 Thanks @Conaclos! - Implemented #7174. useConst no longer reports variables that are read before being written.

    Previously, useConst reported uninitialised variables that were read in an inner function before being written, as shown in the following example:

    let v;
    function f() {
      return v;
    }
    v = 0;

    This can produce false positives in the case where f is called before v has been written, as in the following code:

    let v;
    function f() {
      return v;
    }
    console.log(f()); // print `undefined`
    v = 0;

    Although this is an expected behavior of the original implementation, we consider it problematic since the rule’s fix is marked as safe. To avoid false positives like this, the rule now ignores the previous examples. However, this has the disadvantage of resulting in false negatives, such as not reporting the first example.

Patch Changes

  • #7156 137d111 Thanks @ematipico! - Fixed #7152. Now the rule noDuplicateFontNames correctly detects font names with spaces e.g. Liberation Mono. The diagnostic of the rule now points to the first instances of the repeated font.

    The following example doesn’t trigger the rule anymore:

    c {
      font-family:
        SF Mono,
        Liberation Mono,
        sans-serif;
    }
    d {
      font:
        1em SF Mono,
        Liberation Mono,
        sans-serif;
    }

  • #6907 7331bb9 Thanks @ematipico! - Added a new experimental option that allows parsing of .html files that contain interpolation syntax.

    biome.json
    {
      html: {
        // This is the new, experimental option.
        parser: {
          interpolation: true,
        },
      },
    }
    <h1>{{ $title }}</h1>

  • #7124 3f436b8 Thanks @Jayllyz! - Added the rule useMaxParams.

    This rule enforces a maximum number of parameters for functions to improve code readability and maintainability. Functions with many parameters are difficult to read, understand, and maintain because they require memorizing parameter order and types.

    // Invalid - too many parameters (default max: 4)
    function processData(
      name,
      age,
      email,
      phone,
      address,
      city,
      country,
      zipCode,
    ) {
      // ...
    }
    

    // Valid - within parameter limit
    function processData(userData) {
      const { name, age, email, phone, address, city, country, zipCode } =
        userData;
      // ...
    }
    

    function calculateSum(a, b, c) {
      return a + b + c;
    }

  • #7161 1a14a59 Thanks @ematipico! - Fixed #7160. Now Biome correctly computes ignored files when using formatter.includes, linter.includes and assist.includes inside nested configurations that use "extends": "//".

  • #7081 a081bbe Thanks @Jayllyz! - Added the rule noNextAsyncClientComponent.

    This rule prevents the use of async functions for client components in Next.js applications. Client components marked with “use client” directive should not be async as this can cause hydration mismatches, break component rendering lifecycle, and lead to unexpected behavior with React’s concurrent features.

    "use client";
    

    // Invalid - async client component
    export default async function MyComponent() {
      return <div>Hello</div>;
    }
    

    // Valid - synchronous client component
    export default function MyComponent() {
      return <div>Hello</div>;
    }

  • #7171 5241690 Thanks @siketyan! - Fixed #7162: The noUndeclaredDependencies rule now considers a type-only import as a dev dependency.

    For example, the following code is no longer reported:

    package.json:

    {
      "devDependencies": {
        "type-fest": "*"
      }
    }

    foo.ts:

    import type { SetRequired } from "type-fest";

    Note that you still need to declare the package in the devDependencies section in package.json.

2.2.2

Patch Changes

  • #7266 b270bb5 Thanks @ematipico! - Fixed an issue where Biome got stuck when analyzing some files. This is usually caused by a bug in the inference engine. Now Biome has some guards in place in case the number of types grows too much, and if that happens, a diagnostic is emitted and the inference is halted.

  • #7281 6436180 Thanks @ematipico! - Fixed an issue where the function scanProject wouldn’t work as expected.

  • #7285 1511d0c Thanks @rriski! - Partially fixed #6782: JSX node kinds are now supported in GritQL AST nodes.

  • #7249 dff85c0 Thanks @ematipico! - Fixed #748, where Biome Language Server didn’t show the unsafe fixes when requesting the quick fixes. Now all LSP editors will show also opt-in, unsafe fixes.

  • #7266 b270bb5 Thanks @ematipico! - Fixed #7020: Resolved an issue with analysing types of static member expressions involving unions. If the object type was a union that referenced nested unions, it would trigger an infinite loop as it tried to keep expanding nested unions, and the set of types would grow indefinitely.

  • #7209 679b70e Thanks @patrickshipe! - Resolved an overcorrection in useImportExtensions when importing explicit index files.

    Imports that explicitly reference an index file are now preserved and no longer rewritten to nested index paths.

    Example

    // Before
          import "./sub/index";
          import "./sub/index/index.js";
    

    // After
          import "./sub/index";
          import "./sub/index.js";

  • #7270 953f9c6 Thanks @arendjr! - Fixed #6172: Resolved an issue with inferring types for rest parameters. This issue caused rest-parameter types to be incorrect, and in some cases caused extreme performance regressions in files that contained many methods with rest-parameter definitions.

  • #7234 b7aa111 Thanks @JeetuSuthar! - Fixed #7233: The useIndexOf rule now correctly suggests using indexOf() instead of findIndex().

    The diagnostic message was incorrectly recommending Array#findIndex() over Array#indexOf(), when it should recommend the opposite for simple equality checks.

  • #7283 0b07f45 Thanks @ematipico! - Fixed #7236. Now Biome correctly migrates JSONC configuration files when they are passed using --config-path.

  • #7239 1d643d8 Thanks @minht11! - Fixed an issue where Svelte globals ($state and so on) were not properly recognized inside .svelte.test.ts/js and .svelte.spec.ts/js files.

  • #7264 62fdbc8 Thanks @ematipico! - Fixed a regression where when using --log-kind-pretty wasn’t working anymore as expected.

  • #7244 660031b Thanks @JeetuSuthar! - Fixed #7225: The noExtraBooleanCast rule now preserves parentheses when removing Boolean calls inside negations.

    // Before
    !Boolean(b0 && b1);
    // After
    !(b0 && b1); // instead of !b0 && b1

  • #7298 46a8e93 Thanks @unvalley! - Fixed #6695: useNamingConvention now correctly reports TypeScript parameter properties with modifiers.

    Previously, constructor parameter properties with modifiers like private or readonly were not checked against naming conventions. These properties are now treated consistently with regular class properties.

2.2.3

Patch Changes

  • #7353 4d2b719 Thanks @JeetuSuthar! - Fixed #7340: The linter now allows the navigation property for view-transition in CSS.

    Previously, the linter incorrectly flagged navigation: auto as an unknown property. This fix adds navigation to the list of known CSS properties, following the CSS View Transitions spec.

  • #7275 560de1b Thanks @arendjr! - Fixed #7268: Files that are explicitly passed as CLI arguments are now correctly ignored if they reside in an ignored folder.

  • #7358 963a246 Thanks @ematipico! - Fixed #7085, now the rule noDescendingSpecificity correctly calculates the specificity of selectors when they are included inside a media query.

  • #7387 923674d Thanks @qraqras! - Fixed #7381, now the useOptionalChain rule recognizes optional chaining using Yoda expressions (e.g., undefined !== foo && foo.bar).

  • #7316 f9636d5 Thanks @Conaclos! - Fixed #7289. The rule useImportType now inlines import type into import { type } when the style option is set to inlineType.

    Example:

    import type { T } from "mod";
    // becomes
    import { type T } from "mod";

  • #7350 bb4d407 Thanks @siketyan! - Fixed #7261: two characters ・ (KATAKANA MIDDLE DOT, U+30FB) and ･ (HALFWIDTH KATAKANA MIDDLE DOT, U+FF65) are no longer considered as valid characters in identifiers. Property keys containing these character(s) are now preserved as string literals.

  • #7377 811f47b Thanks @ematipico! - Fixed a bug where the Biome Language Server didn’t correctly compute the diagnostics of a monorepo setting, caused by an incorrect handling of the project status.

  • #7245 fad34b9 Thanks @kedevked! - Added the new lint rule useConsistentArrowReturn.

    This rule enforces a consistent return style for arrow functions.

    Invalid

    const f = () => {
      return 1;
    };

    This rule is a port of ESLint’s arrow-body-style rule.

  • #7370 e8032dd Thanks @fireairforce! - Support dynamic import defer and import source. The syntax looks like:

    import.source("foo");
    import.source("x", { with: { attr: "val" } });
    import.defer("foo");
    import.defer("x", { with: { attr: "val" } });

  • #7369 b1f8cbd Thanks @siketyan! - Range suppressions are now supported for Grit plugins.

    For JavaScript, you can suppress a plugin as follows:

    // biome-ignore-start lint/plugin/preferObjectSpread: reason
    Object.assign({ foo: "bar" }, baz);
    // biome-ignore-end lint/plugin/preferObjectSpread: reason

    For CSS, you can suppress a plugin as follows:

    body {
      /* biome-ignore-start lint/plugin/useLowercaseColors: reason */
      color: #fff;
      /* biome-ignore-end lint/plugin/useLowercaseColors: reason */
    }

  • #7384 099507e Thanks @ematipico! - Reduced the severity of certain diagnostics emitted when Biome deserializes the configuration files. Now these diagnostics are emitted as Information severity, which means that they won’t interfere when running commands with --error-on-warnings

  • #7302 2af2380 Thanks @unvalley! - Fixed #7301: useReadonlyClassProperties now correctly skips JavaScript files.

  • #7288 94d85f8 Thanks @ThiefMaster! - Fixed #7286. Files are now formatted with JSX behavior when javascript.parser.jsxEverywhere is explicitly set.

    Previously, this flag was only used for parsing, but not for formatting, which resulted in incorrect formatting of conditional expressions when JSX syntax is used in .js files.

  • #7311 62154b9 Thanks @qraqras! - Added the new nursery rule noUselessCatchBinding. This rule disallows unnecessary catch bindings.

    try {
        // Do something
     } catch (unused) {}
     } catch {}

  • #7349 45c1dfe Thanks @ematipico! - Fixed #4298. Biome now correctly formats CSS declarations when it contains one single value:

    .bar {
      --123456789012345678901234567890: var(--1234567890123456789012345678901234567);
      --123456789012345678901234567890: var(
        --1234567890123456789012345678901234567
      );
    }

  • #7295 7638e84 Thanks @ematipico! - Fixed #7130. Removed the emission of a false-positive diagnostic. Biome no longer emits the following diagnostic:

    lib/main.ts:1:5 suppressions/unused ━━━━━━━━━━━━━━━━━━━━━━━━━
    

      ⚠ Suppression comment has no effect because the tool is not enabled.
    

      > 1 │ /** biome-ignore-all assist/source/organizeImports: For the lib root file, we don't want to organize exports */
          │     ^^^^^^^^^^^^^^^^

  • #7377 811f47b Thanks @ematipico! - Fixed #7371 where the Biome Language Server didn’t correctly recompute the diagnostics when updating a nested configuration file.

  • #7348 ac27fc5 Thanks @ematipico! - Fixed #7079. Now the rule useSemanticElements doesn’t trigger components and custom elements.

  • #7389 ab06a7e Thanks @Conaclos! - Fixed #7344. useNamingConvention no longer reports interfaces defined in global declarations.

    Interfaces declared in global declarations augment existing interfaces. Thus, they must be ignored.

    In the following example, useNamingConvention reported HTMLElement. It is now ignored.

    export {};
    declare global {
      interface HTMLElement {
        foo(): void;
      }
    }

  • #7315 4a2bd2f Thanks @vladimir-ivanov! - Fixed #7310: useReadonlyClassProperties correctly handles nested assignments, avoiding false positives when a class property is assigned within another assignment expression.

    Example of code that previously triggered a false positive but is now correctly ignored:

    class test {
      private thing: number = 0; // incorrectly flagged
    

      public incrementThing(): void {
        const temp = { x: 0 };
        temp.x = this.thing++;
      }
    }

2.2.4

Patch Changes

  • #7453 aa8cea3 Thanks @arendjr! - Fixed #7242: Aliases specified in package.json’s imports section now support having multiple targets as part of an array.

  • #7454 ac17183 Thanks @arendjr! - Greatly improved performance of noImportCycles by eliminating allocations.

    In one repository, the total runtime of Biome with only noImportCycles enabled went from ~23s down to ~4s.

  • #7447 7139aad Thanks @rriski! - Fixes #7446. The GritQL $... spread metavariable now correctly matches members in object literals, aligning its behavior with arrays and function calls.

  • #6710 98cf9af Thanks @arendjr! - Fixed #4723: Type inference now recognises index signatures and their accesses when they are being indexed as a string.

    Example

    type BagOfPromises = {
      // This is an index signature definition. It declares that instances of type
      // `BagOfPromises` can be indexed using arbitrary strings.
      [property: string]: Promise<void>;
    };
    

    let bag: BagOfPromises = {};
    // Because `bag.iAmAPromise` is equivalent to `bag["iAmAPromise"]`, this is
    // considered an access to the string index, and a Promise is expected.
    bag.iAmAPromise;

  • #7415 d042f18 Thanks @qraqras! - Fixed #7212, now the useOptionalChain rule recognizes optional chaining using typeof (e.g., typeof foo !== 'undefined' && foo.bar).

  • #7419 576baf4 Thanks @Conaclos! - Fixed #7323. noUnusedPrivateClassMembers no longer reports as unused TypeScript private members if the rule encounters a computed access on this.

    In the following example, member as previously reported as unused. It is no longer reported.

    class TsBioo {
      private member: number;
    

      set_with_name(name: string, value: number) {
        this[name] = value;
      }
    }

  • 351bccd Thanks @ematipico! - Added the new nursery lint rule noJsxLiterals, which disallows the use of string literals inside JSX.

    The rule catches these cases:

    <>
      <div>test</div> {/* test is invalid */}
      <>test</>
      <div>
        {/* this string is invalid */}
        asdjfl test foo
      </div>
    </>

  • #7406 b906112 Thanks @mdevils! - Fixed an issue (#6393) where the useHookAtTopLevel rule reported excessive diagnostics for nested hook calls.

    The rule now reports only the offending top-level call site, not sub-hooks of composite hooks.

    // Before: reported twice (useFoo and useBar).
    function useFoo() {
      return useBar();
    }
    function Component() {
      if (cond) useFoo();
    }
    // After: reported once at the call to useFoo().

  • #7461 ea585a9 Thanks @arendjr! - Improved performance of noPrivateImports by eliminating allocations.

    In one repository, the total runtime of Biome with only noPrivateImports enabled went from ~3.2s down to ~1.4s.

  • 351bccd Thanks @ematipico! - Fixed #7411. The Biome Language Server had a regression where opening an editor with a file already open wouldn’t load the project settings correctly.

  • #7142 53ff5ae Thanks @Netail! - Added the new nursery rule noDuplicateDependencies, which verifies that no dependencies are duplicated between the bundledDependencies, bundleDependencies, dependencies, devDependencies, overrides, optionalDependencies, and peerDependencies sections.

    For example, the following snippets will trigger the rule:

    {
      "dependencies": {
        "foo": ""
      },
      "devDependencies": {
        "foo": ""
      }
    }
    {
      "dependencies": {
        "foo": ""
      },
      "optionalDependencies": {
        "foo": ""
      }
    }
    {
      "dependencies": {
        "foo": ""
      },
      "peerDependencies": {
        "foo": ""
      }
    }

  • 351bccd Thanks @ematipico! - Fixed #3824. Now the option CLI --color is correctly applied to logging too.

2.2.5

Patch Changes

  • #7597 5c3d542 Thanks @arendjr! - Fixed #6432: useImportExtensions now works correctly with aliased paths.

  • #7269 f18dac1 Thanks @CDGardner! - Fixed #6648, where Biome’s noUselessFragments contained inconsistencies with ESLint for fragments only containing text.

    Previously, Biome would report that fragments with only text were unnecessary under the noUselessFragments rule. Further analysis of ESLint’s behavior towards these cases revealed that text-only fragments (<>A</a>, <React.Fragment>B</React.Fragment>, <RenamedFragment>B</RenamedFragment>) would not have noUselessFragments emitted for them.

    On the Biome side, instances such as these would emit noUselessFragments, and applying the suggested fix would turn the text content into a proper JS string.

    // Ended up as: - const t = "Text"
    const t = <>Text</>
    

    // Ended up as: - const e = t ? "Option A" : "Option B"
    const e = t ? <>Option A</> : <>Option B</>
    

    /* Ended up as:
      function someFunc() {
        return "Content desired to be a multi-line block of text."
      }
    */
    function someFunc() {
      return <>
        Content desired to be a multi-line
        block of text.
      <>
    }

    The proposed update was to align Biome’s reaction to this rule with ESLint’s; the aforementioned examples will now be supported from Biome’s perspective, thus valid use of fragments.

    // These instances are now valid and won't be called out by noUselessFragments.
    

    const t = <>Text</>
    const e = t ? <>Option A</> : <>Option B</>
    

    function someFunc() {
      return <>
        Content desired to be a multi-line
        block of text.
      <>
    }

  • #7498 002cded Thanks @siketyan! - Fixed #6893: The useExhaustiveDependencies rule now correctly adds a dependency that is captured in a shorthand object member. For example:

    useEffect(() => {
      console.log({ firstId, secondId });
    }, []);

    is now correctly fixed to:

    useEffect(() => {
      console.log({ firstId, secondId });
    }, [firstId, secondId]);

  • #7509 1b61631 Thanks @siketyan! - Added a new lint rule noReactForwardRef, which detects usages of forwardRef that is no longer needed and deprecated in React 19.

    For example:

    export const Component = forwardRef(function Component(props, ref) {
      return <div ref={ref} />;
    });

    will be fixed to:

    export const Component = function Component({ ref, ...props }) {
      return <div ref={ref} />;
    };

    Note that the rule provides an unsafe fix, which may break the code. Don’t forget to review the code after applying the fix.

  • #7520 3f06e19 Thanks @arendjr! - Added new nursery rule noDeprecatedImports to flag imports of deprecated symbols.

    Invalid example

    foo.js
    import { oldUtility } from "./utils.js";
    utils.js
    /**
     * @deprecated
     */
    export function oldUtility() {}

    Valid examples

    foo.js
    import { newUtility, oldUtility } from "./utils.js";
    utils.js
    export function newUtility() {}
    

    // @deprecated (this is not a JSDoc comment)
    export function oldUtility() {}

  • #7457 9637f93 Thanks @kedevked! - Added style and requireForObjectLiteral options to the lint rule useConsistentArrowReturn.

    This rule enforces a consistent return style for arrow functions. It can be configured with the following options:

    • style: (default: asNeeded)
      • always: enforces that arrow functions always have a block body.
      • never: enforces that arrow functions never have a block body, when possible.
      • asNeeded: enforces that arrow functions have a block body only when necessary (e.g. for object literals).

    style: "always"

    Invalid:

    const f = () => 1;

    Valid:

    const f = () => {
      return 1;
    };

    style: "never"

    Invalid:

    const f = () => {
      return 1;
    };

    Valid:

    const f = () => 1;

    style: "asNeeded"

    Invalid:

    const f = () => {
      return 1;
    };

    Valid:

    const f = () => 1;

    style: "asNeeded" and requireForObjectLiteral: true

    Valid:

    const f = () => {
      return { a: 1 };
    };

  • #7510 527cec2 Thanks @rriski! - Implements #7339. GritQL patterns can now use native Biome AST nodes using their PascalCase names, in addition to the existing TreeSitter-compatible snake_case names.

    engine biome(1.0)
    language js(typescript,jsx)
    

    or {
      // TreeSitter-compatible pattern
      if_statement(),
    

      // Native Biome AST node pattern
      JsIfStatement()
    } as $stmt where {
      register_diagnostic(
        span=$stmt,
        message="Found an if statement"
      )
    }

  • #7574 47907e7 Thanks @kedevked! - Fixed 7574. The diagnostic message for the rule useSolidForComponent now correctly emphasizes <For /> and provides a working hyperlink to the Solid documentation.

  • #7497 bd70f40 Thanks @siketyan! - Fixed #7320: The useConsistentCurlyBraces rule now correctly detects a string literal including " inside a JSX attribute value.

  • #7522 1af9931 Thanks @Netail! - Added extra references to external rules to improve migration for the following rules: noUselessFragments & noNestedComponentDefinitions

  • #7597 5c3d542 Thanks @arendjr! - Fixed an issue where package.json manifests would not be correctly discovered when evaluating files in the same directory.

  • #7565 38d2098 Thanks @siketyan! - The resolver can now correctly resolve .ts, .tsx, .d.ts, .js files by .js extension if exists, based on the file extension substitution in TypeScript.

    For example, the linter can now detect the floating promise in the following situation, if you have enabled the noFloatingPromises rule.

    foo.ts

    export async function doSomething(): Promise<void> {}

    bar.ts

    import { doSomething } from "./foo.js"; // doesn't exist actually, but it is resolved to `foo.ts`
    

    doSomething(); // floating promise!

  • #7542 cadad2c Thanks @mdevils! - Added the rule noVueDuplicateKeys, which prevents duplicate keys in Vue component definitions.

    This rule prevents the use of duplicate keys across different Vue component options such as props, data, computed, methods, and setup. Even if keys don’t conflict in the script tag, they may cause issues in the template since Vue allows direct access to these keys.

    Invalid examples
    <script>
    export default {
      props: ["foo"],
      data() {
        return {
          foo: "bar",
        };
      },
    };
    </script>
    <script>
    export default {
      data() {
        return {
          message: "hello",
        };
      },
      methods: {
        message() {
          console.log("duplicate key");
        },
      },
    };
    </script>
    <script>
    export default {
      computed: {
        count() {
          return this.value * 2;
        },
      },
      methods: {
        count() {
          this.value++;
        },
      },
    };
    </script>
    Valid examples
    <script>
    export default {
      props: ["foo"],
      data() {
        return {
          bar: "baz",
        };
      },
      methods: {
        handleClick() {
          console.log("unique key");
        },
      },
    };
    </script>
    <script>
    export default {
      computed: {
        displayMessage() {
          return this.message.toUpperCase();
        },
      },
      methods: {
        clearMessage() {
          this.message = "";
        },
      },
    };
    </script>

  • #7546 a683acc Thanks @siketyan! - Internal data for Unicode strings have been updated to Unicode 17.0.

  • #7497 bd70f40 Thanks @siketyan! - Fixed #7256: The useConsistentCurlyBraces rule now correctly ignores a string literal with braces that contains only whitespaces. Previously, literals that contains single whitespace were only allowed.

  • #7565 38d2098 Thanks @siketyan! - The useImportExtensions rule now correctly detects imports with an invalid extension. For example, importing .ts file with .js extension is flagged by default. If you are using TypeScript with neither the allowImportingTsExtensions option nor the rewriteRelativeImportExtensions option, it’s recommended to turn on the forceJsExtensions option of the rule.

  • #7581 8653921 Thanks @lucasweng! - Fixed #7470: solved a false positive for noDuplicateProperties. Previously, declarations in @container and @starting-style at-rules were incorrectly flagged as duplicates of identical declarations at the root selector.

    For example, the linter no longer flags the display declaration in @container or the opacity declaration in @starting-style.

    a {
      display: block;
      @container (min-width: 600px) {
        display: none;
      }
    }
    

    [popover]:popover-open {
      opacity: 1;
      @starting-style {
        opacity: 0;
      }
    }

  • #7529 fea905f Thanks @qraqras! - Fixed #7517: the useOptionalChain rule no longer suggests changes for typeof checks on global objects.

    // ok
    typeof window !== "undefined" && window.location;

  • #7476 c015765 Thanks @ematipico! - Fixed a bug where the suppression action for noPositiveTabindex didn’t place the suppression comment in the correct position.

  • #7511 a0039fd Thanks @arendjr! - Added nursery rule noUnusedExpressions to flag expressions used as a statement that is neither an assignment nor a function call.

    Invalid examples

    f; // intended to call `f()` instead
    function foo() {
      0; // intended to `return 0` instead
    }

    Valid examples

    f();
    function foo() {
      return 0;
    }

  • #7564 40e515f Thanks @turbocrime! - Fixed #6617: improved useIterableCallbackReturn to correctly handle arrow functions with a single-expression void body.

    Now the following code doesn’t trigger the rule anymore:

    [].forEach(() => void null);

2.2.6

Patch Changes

  • #7071 a8e7301 Thanks @ptkagori! - Added the useQwikMethodUsage lint rule for the Qwik domain.

    This rule validates Qwik hook usage. Identifiers matching useXxx must be called only within serialisable reactive contexts (for example, inside component$, route loaders/actions, or within other Qwik hooks), preventing common Qwik antipatterns.

    Invalid:

    // Top-level hook call is invalid.
    const state = useStore({ count: 0 });
    

    function helper() {
      // Calling a hook in a non-reactive function is invalid.
      const loc = useLocation();
    }

    Valid:

    component$(() => {
      const state = useStore({ count: 0 }); // OK inside component$.
      return <div>{state.count}</div>;
    });
    

    const handler = $(() => {
      const loc = useLocation(); // OK inside a $-wrapped closure.
      console.log(loc.params);
    });

  • #7685 52071f5 Thanks @denbezrukov! - Fixed #6981: The NoUnknownPseudoClass rule no longer reports local pseudo-classes when CSS Modules are used.

  • #7640 899f7b2 Thanks @arendjr! - Fixed #7638: useImportExtensions no longer emits diagnostics on valid import paths that end with a query or hash.

    Example

    // This no longer warns if `index.css` exists:
    import style from "../theme/index.css?inline";

  • #7071 a8e7301 Thanks @ptkagori! - Added the useQwikValidLexicalScope rule to the Qwik domain.

    This rule helps you avoid common bugs in Qwik components by checking that your variables and functions are declared in the correct place.

    Invalid:

    // Invalid: state defined outside the component's lexical scope.
    let state = useStore({ count: 0 });
    const Component = component$(() => {
      return (
        <button onClick$={() => state.count++}>Invalid: {state.count}</button>
      );
    });

    Valid:

    // Valid: state initialised within the component's lexical scope and captured by the event.
    const Component = component$(() => {
      const state = useStore({ count: 0 });
      return <button onClick$={() => state.count++}>Valid: {state.count}</button>;
    });

  • #7620 5beb1ee Thanks @Netail! - Added the rule useDeprecatedDate, which makes a deprecation date required for the graphql @deprecated directive.

    Invalid
    query {
      member @deprecated(reason: "Use `members` instead") {
        id
      }
    }
    Valid
    query {
      member
        @deprecated(reason: "Use `members` instead", deletionDate: "2099-12-25") {
        id
      }
    }

  • #7709 d6da4d5 Thanks @siketyan! - Fixed #7704: The useExhaustiveDependencies rule now correctly adds an object dependency when its method is called within the closure.

    For example:

    function Component(props) {
      useEffect(() => {
        props.foo();
      }, []);
    }

    will now be fixed to:

    function Component(props) {
      useEffect(() => {
        props.foo();
      }, [props]);
    }

  • #7624 309ae41 Thanks @lucasweng! - Fixed #7595: noUselessEscapeInString no longer reports $\{ escape in template literals.

  • #7665 29e4229 Thanks @ryan-m-walker! - Fixed #7619: Added support for parsing the CSS :state() pseudo-class.

    custom-selector:state(checked) {
    }

  • #7608 41df59b Thanks @ritoban23! - Fixed #7604: the useMaxParams rule now highlights parameter lists instead of entire function bodies. This provides more precise error highlighting. Previously, the entire function was highlighted; now only the parameter list is highlighted, such as (a, b, c, d, e, f, g, h).

  • #7643 459a6ac Thanks @daivinhtran! - Fixed #7580: Include plugin in summary report

2.2.7

Patch Changes

2.3.0

Minor Changes

  • #7263 a3e3369 Thanks @arendjr! - Biome’s resolver now supports baseUrl if specified in tsconfig.json.

    Example

    Given the following file structure:

    tsconfig.json

    {
      "compilerOptions": {
        "baseUrl": "./src"
      }
    }

    src/foo.ts

    export function foo() {}

    In this scenario, import { foo } from "foo"; should work regardless of the location of the file containing the import statement.

    Fixes #6432.

  • #7745 6fcbc07 Thanks @dyc3! - Added ignore option to noUnknownAtRules. If an unknown at-rule matches any of the items provided in ignore, a diagnostic won’t be emitted.

  • #7753 63cb7ff Thanks @ematipico! - Enhanced the init command. The init command now checks if the existing project contains known ignore files and known generated folders.

    If Biome finds .gitignore or .ignore files, it will add the following configuration to biome.json:

    {
      "vcs": {
        "enabled": true,
        "clientKind": "git",
        "useIgnoreFile": true
      }
    }

    If Biome finds a dist/ folder, it will exclude it automatically using the double-exclude syntax:

    {
      "files": {
        "includes": ["**", "!!**/dist"]
      }
    }

  • #7548 85d3a3a Thanks @siketyan! - The rules in a domain are no longer enabled automatically by the installed dependencies unless the rule is recommended.

  • #7723 d3aac63 Thanks @ematipico! - Added --css-parse-css-modules CLI flag to control whether CSS Modules syntax is enabled.

    You can now enable or disable CSS Modules parsing directly from the command line:

    Terminal window
    biome check --css-parse-css-modules=true file.module.css
    biome format --css-parse-css-modules=true file.module.css
    biome lint --css-parse-css-modules=true file.module.css
    biome ci --css-parse-css-modules=true file.module.css

  • #7723 d3aac63 Thanks @ematipico! - Added --css-parse-tailwind-directives CLI flag to control whether Tailwind CSS 4.0 directives and functions are enabled.

    You can now enable or disable Tailwind CSS 4.0 directive parsing directly from the command line:

    Terminal window
    biome check --css-parse-tailwind-directives=true file.css
    biome format --css-parse-tailwind-directives=true file.css
    biome lint --css-parse-tailwind-directives=true file.css
    biome ci --css-parse-tailwind-directives=true file.css

  • #7330 272632f Thanks @ematipico! - Updated the formatting of .svelte and .vue files. Now the indentation of the JavaScript blocks matches Prettier’s:

    <script>
     import Component from "./Component"
       import Component from "./Component"
    </script>

  • #7333 de0d2d6 Thanks @dyc3! - Implemented the indentScriptAndStyle option for vue and svelte files, with the default set to false to match Prettier’s vueIndentScriptAndStyle option. When enabled, this option indents the content within <script> and <style> tags to align with the surrounding HTML structure.

    It can be enabled with this configuration:

    {
      "html": {
        "formatter": {
          "indentScriptAndStyle": true
        }
      }
    }

    Which will format this code to:

    <script>
    import Component from "./Component.vue";
    </script>

  • #7359 ebbddc4 Thanks @arendjr! - Deprecated the option files.experimentalScannerIgnores in favour of force-ignore syntax in files.includes.

    files.includes supports ignoring files by prefixing globs with an exclamation mark (!). With this change, it also supports force-ignoring globs by prefixing them with a double exclamation mark (!!).

    The effect of force-ignoring is that the scanner will not index files matching the glob, even in project mode, even if those files are imported by other files, and even if they are files that receive special treatment by Biome, such as nested biome.json files.

    Example

    Let’s take the following configuration:

    {
      "files": {
        "includes": [
          "**",
          "!**/generated",
          "!!**/dist",
          "fixtures/example/dist/*.js"
        ]
      },
      "linter": {
        "domains": {
          "project": "all"
        }
      }
    }

    This configuration achieves the following:

    • Because the project domain is enabled, all supported files in the project are indexed and processed by the linter, except:
    • Files inside a generated folder are not processed by the linter, but they will get indexed if a file outside a generated folder imports them.
    • Files inside a dist folder are never indexed nor processed, not even if they are imported for any purpose, except:
    • When the dist folder is inside fixtures/example/, its .js files do get both indexed and processed.

    In general, we now recommend using the force-ignore syntax for any folders that contain output files, such as build/ and dist/. For such folders, it is highly unlikely that indexing has any useful benefits. For folders containing generated files, you may wish to use the regular ignore syntax so that type information can still be extracted from the files.

    experimentalScannerIgnores will continue to work for now, but you’ll see a deprecation warning if you still use it.

    Run the biome migrate --write command to automatically update the configuration file.

  • #7698 3b6f5e3 Thanks @ematipico! - Added a new reporter named rdjson. This reporter prints diagnostics following the RDJSON format:

    The following command:

    Terminal window
    biome check --reporter=rdjson

    Will emit diagnostics in the following format:

    {
      "source": {
        "name": "Biome",
        "url": "https://biomejs.dev"
      },
      "diagnostics": [
        {
          "code": {
            "url": "https://biomejs.dev/linter/rules/no-unused-imports",
            "value": "lint/correctness/noUnusedImports"
          },
          "location": {
            "path": "index.ts",
            "range": {
              "end": {
                "column": 11,
                "line": 0
              },
              "start": {
                "column": 7,
                "line": 0
              }
            }
          },
          "message": "This import is unused."
        },
        {
          "code": {
            "url": "https://biomejs.dev/linter/rules/no-unused-imports",
            "value": "lint/correctness/noUnusedImports"
          },
          "location": {
            "path": "index.ts",
            "range": {
              "end": {
                "column": 10,
                "line": 1
              },
              "start": {
                "column": 9,
                "line": 1
              }
            }
          },
          "message": "Several of these imports are unused."
        }
      ]
    }

  • #7719 188a767 Thanks @cadunass! - The formatWithErrors option can now be set via CLI using the --format-with-errors flag.

    This flag was previously only available in the configuration file. It allows formatting to proceed on files with syntax errors, which is useful during development when you want to auto-format code while fixing syntax issues.

    Example

    Terminal window
    biome format --format-with-errors=true --write file.js

  • #7723 d3aac63 Thanks @ematipico! - Added --json-parse-allow-comments CLI flag to control whether comments are allowed in JSON files.

    You can now enable or disable comment parsing in JSON files directly from the command line:

    Terminal window
    biome check --json-parse-allow-comments=true file.json
    biome format --json-parse-allow-comments=true file.json
    biome lint --json-parse-allow-comments=true file.json
    biome ci --json-parse-allow-comments=true file.json

  • #7723 d3aac63 Thanks @ematipico! - Added --json-parse-allow-trailing-commas CLI flag to control whether trailing commas are allowed in JSON files.

    You can now enable or disable trailing comma parsing in JSON files directly from the command line:

    Terminal window
    biome check --json-parse-allow-trailing-commas=true file.json
    biome format --json-parse-allow-trailing-commas=true file.json
    biome lint --json-parse-allow-trailing-commas=true file.json
    biome ci --json-parse-allow-trailing-commas=true file.json

  • #7758 cea002f Thanks @ematipico! - Promoted new lint rules:

    • Promoted noNonNullAssertedOptionalChain to the suspicious group
    • Promoted useReactFunctionComponents to the style group
    • Promoted useImageSize to the correctness group
    • Promoted useConsistentTypeDefinitions to the style group
    • Promoted useQwikClasslist to the correctness group
    • Promoted noSecrets to the security group

    Removed the lint rule useAnchorHref, because its use case is covered by useValidAnchor.

  • #6356 296627d Thanks @wrick17! - Added the new checkstyle reporter. When --reporter=checkstyle is passed to the CLI, Biome will emit diagnostics for Checkstyle format:

    <?xml version="1.0" encoding="utf-8"?>
    <checkstyle version="4.3">
      <file name="index.ts">
        <error line="1" column="8" severity="warning" message="This import is unused." source="lint/correctness/noUnusedImports" />
        <error line="2" column="10" severity="warning" message="Several of these imports are unused." source="lint/correctness/noUnusedImports" />
        <error line="8" column="5" severity="warning" message="This variable f is unused." source="lint/correctness/noUnusedVariables" />
        <error line="9" column="7" severity="warning" message="This variable f is unused." source="lint/correctness/noUnusedVariables" />
        <error line="1" column="1" severity="error" message="The imports and exports are not sorted." source="assist/source/organizeImports" />
        <error line="4" column="3" severity="error" message="Using == may be unsafe if you are relying on type coercion." source="lint/suspicious/noDoubleEquals" />
        <error line="6" column="1" severity="error" message="This is an unexpected use of the debugger statement." source="lint/suspicious/noDebugger" />
        <error line="8" column="5" severity="error" message="This variable implicitly has the any type." source="lint/suspicious/noImplicitAnyLet" />
        <error line="9" column="7" severity="error" message="This variable implicitly has the any type." source="lint/suspicious/noImplicitAnyLet" />
        <error line="2" column="10" severity="error" message="Shouldn&apos;t redeclare &apos;z&apos;. Consider to delete it or rename it." source="lint/suspicious/noRedeclare" />
        <error line="9" column="7" severity="error" message="Shouldn&apos;t redeclare &apos;f&apos;. Consider to delete it or rename it." source="lint/suspicious/noRedeclare" />
        <error line="0" column="0" severity="error" message="Formatter would have printed the following content:" source="format" />
      </file>
      <file name="main.ts">
        <error line="1" column="8" severity="warning" message="This import is unused." source="lint/correctness/noUnusedImports" />
        <error line="2" column="10" severity="warning" message="Several of these imports are unused." source="lint/correctness/noUnusedImports" />
        <error line="8" column="5" severity="warning" message="This variable f is unused." source="lint/correctness/noUnusedVariables" />
        <error line="9" column="7" severity="warning" message="This variable f is unused." source="lint/correctness/noUnusedVariables" />
        <error line="1" column="1" severity="error" message="The imports and exports are not sorted." source="assist/source/organizeImports" />
        <error line="4" column="3" severity="error" message="Using == may be unsafe if you are relying on type coercion." source="lint/suspicious/noDoubleEquals" />
        <error line="6" column="1" severity="error" message="This is an unexpected use of the debugger statement." source="lint/suspicious/noDebugger" />
        <error line="8" column="5" severity="error" message="This variable implicitly has the any type." source="lint/suspicious/noImplicitAnyLet" />
        <error line="9" column="7" severity="error" message="This variable implicitly has the any type." source="lint/suspicious/noImplicitAnyLet" />
        <error line="2" column="10" severity="error" message="Shouldn&apos;t redeclare &apos;z&apos;. Consider to delete it or rename it." source="lint/suspicious/noRedeclare" />
        <error line="9" column="7" severity="error" message="Shouldn&apos;t redeclare &apos;f&apos;. Consider to delete it or rename it." source="lint/suspicious/noRedeclare" />
        <error line="0" column="0" severity="error" message="Formatter would have printed the following content:" source="format" />
      </file>
    </checkstyle>

  • #7488 b13e524 Thanks @kpapa05! - Added “@rbxts/react” as an alias for “react” for handling the reactClassic jsxRuntime.

  • #7536 0bccd34 Thanks @TheAlexLichter! - Added .oxlintrc.json to well-known files.

  • #7548 85d3a3a Thanks @siketyan! - The following rules are now a part of the react domain, and they won’t be enabled automatically unless you enabled the domain, or Biome detects react as a dependency of your closest package.json:

  • #7667 480909a Thanks @ematipico! - Added the ability to show severity Information diagnostics in reporter outputs.

    If one or more rules are triggered, and they are configured to emit an Information diagnostic, now they’re counted in the final output:

    Terminal window
    Checked 1 file in <TIME>. No fixes applied.
    Found 1 info.

  • #7702 28e8860 Thanks @ematipico! - Added linting and assist support for .html files, with addition of two new configurations:

    • html.linter.enabled
    • html.assist.enabled

    The HTML linter, in this release, only contains the rule noHeaderScope. More rules will be released in the upcoming releases.

  • #7164 f66b0c5 Thanks @dyc3! - Added a new CSS parser option tailwindDirectives. Enabling this option will allow all of Tailwind v4’s syntax additions to be parsed and formatted by Biome.

    You can enable this by setting css.parser.tailwindDirectives to true in your Biome configuration.

    {
      "css": {
        "parser": {
          "tailwindDirectives": true
        }
      }
    }

  • #7669 6ed4d16 Thanks @barklund! - React 19.2 support is now supported in Biome:

  • #7702 28e8860 Thanks @ematipico! - Added experimental full support for HTML, Vue, Svelte and Astro files. In this release, the HTML parser has been enhanced, and it’s now able to parse .vue, .svelte and .astro files.

    This means that now Biome is able to lint and format the JavaScript (TypeScript), HTML and CSS code that is contained in these files.

    Now that the main architecture is stable and working, in the upcoming patches and minors we will also fix possible inaccuracies and edge cases coming from existing lint rules, such as noUnusedVariables inside <script> blocks or frontmatter.

    The support is considered experimental because there might be cases that aren’t fine-parsed yet, hence causing possible inaccuracies when it comes to formatting and linting.

  • #7599 09445c8 Thanks @anaisbetts! - #### lineEnding has a new option auto

    The option lineEnding now has a variant called auto to match the operating system’s expected line-ending style: on Windows, this will be CRLF (\r\n), and on macOS / Linux, this will be LF (\n).

    This allows for cross-platform projects that use Biome not to have to force one option or the other, which aligns better with Git’s default behavior on these platforms.

    Example usage:

    {
      "formatter": {
        "lineEnding": "auto"
      }
    }
    Terminal window
    biome format --line-ending auto

  • #7392 e4feb8e Thanks @ematipico! - Added new capabilities to the CLI arguments --skip and --only, available to the biome lint command.

    --skip and --only can now accept domain names; when provided, Biome will run or skip all the rules that belong to a certain domain.

    For example, the following command will only run the rules that belong to the next domain:

    Terminal window
    biome lint --only=next

    Another example, the following command will skip the rules that belong to the project domain:

    Terminal window
    biome lint --skip=project

  • #7702 28e8860 Thanks @ematipico! - Added a new option called html.interpolation. This option enables the parsing of text expressions (or interpolation) in HTML files.

    The following file.html will be correctly formatted:

    file.html
    <div>
      Hello {{ name }}!
      <p>Your balance is: {{ account.balance }}</p>
      <button>{{ isLoading ? "Loading..." : "Submit" }}</button>
    </div>

    To note that html.interpolation only parses text expressions that are delimited by double curly braces ({{ }}). The content of expressions is parsed as normal text.

Patch Changes

  • #7712 fcc9b42 Thanks @minht11! - Added new rule useVueDefineMacrosOrder which allows enforcing specific order for Vue compiler macros.

    In this example, the rule will suggest moving defineProps before defineEmits:

    <script lang="ts" setup>
    const emit = defineEmits(["update"]);
    const props = defineProps<{ name: string }>();
    </script>

  • #7698 3b6f5e3 Thanks @ematipico! - Fixed an issue where the JUnit reporter returned a zero-based location. Now the location returned is one-based.

  • #7819 ef45056 Thanks @ematipico! - Fixed #7788. Removes some error logging that were emitted when loading possible configuration files.

  • #7593 e51dd55 Thanks @arendjr! - Fixed an issue with the files.maxSize setting. Previously the setting would always be looked up in the root settings, even in monorepos where a closer biome.json is available. It now correctly uses the nearest configuration.

  • #7825 ad55b35 Thanks @Conaclos! - Fixed #7798. useNamingConvention no longer panics when it encounters a name that consists of a single dollar sign $ that doesn’t match a custom convention.

  • #7764 93be2ab Thanks @gaauwe! - Fixed #6589: Biome now properly loads extension settings before loading the configuration file when opening a text document in the LSP server.

2.3.1

Patch Changes

  • #7840 72afdfa Thanks @ematipico! - Fixed #7838, which caused the new --css-parse-* arguments not being recognised by the ci command.

  • #7789 d5b416e Thanks @fronterior! - Fixed the LSP method workspace/didChangeWorkspaceFolders to perform incremental updates instead of replacing the entire folder list.

  • #7852 bd254c7 Thanks @dyc3! - Fixed #7843: The CSS parser, when tailwindDirectives is enabled, correctly parses --*: initial;.

  • #7872 0fe13fe Thanks @dyc3! - Fixed #7861: The HTML parser will now accept Svelte attribute shorthand syntax in .svelte files.

  • #7866 7b2600b Thanks @dyc3! - Fixed #7860: The css parser, with tailwindDirectives enabled, will now accept @plugin options.

  • #7853 fe90c78 Thanks @dyc3! - Fixed #7848: The css parser with tailwindDirectives enabled will now correctly parse tailwind’s source exclude syntax: @source not "foo.css";

  • #7878 c9f7fe5 Thanks @ematipico! - Fixed #7857: Biome now parses <script> tags as TypeScript when analysing .astro files.

  • #7867 b42b718 Thanks @smorimoto! - Fixed incorrect option name in HTML parser error message.

    The error message for disabled text expressions incorrectly referred to the html.parser.textExpression option, which does not exist. Updated it to reference the correct html.parser.interpolation option.

2.3.2

Patch Changes

  • #7859 c600618 Thanks @Netail! - Added the nursery rule noIncrementDecrement, disallows the usage of the unary operators ++ and —.

  • #7901 0d17b05 Thanks @ematipico! - Fixed #7837, where Biome couldn’t properly parse text expressions that contained nested curly brackets. This was breaking parsing in Astro and Svelte files.

  • #7874 e617d36 Thanks @Bertie690! - Fixed #7230: noUselessStringConcat no longer emits false positives for multi-line strings with leading + operators.

    Previously, the rule did not check for leading newlines on the + operator, emitting false positives if one occurred at the start of a line.
    Notably, formatting with operatorLinebreak="before" would move the + operators to the start of lines automatically, resulting in spurious errors whenever a multi-line string was used.

    Now, the rule correctly detects and ignores multi-line concatenations with leading operators as well, working regardless of the setting of operatorLinebreak.

    Example

    // The following code used to error if the `+` operators were at the start of lines (as opposed to the end).
    // Now, the rule correctly recognizes this as a stylistic concatenation and ignores it.
    const reallyLongStringThatShouldNotError =
      "Lorem ipsum dolor sit amet consectetur adipiscing elit." +
      "Quisque faucibus ex sapien vitae pellentesque sem placerat." +
      "In id cursus mi pretium tellus duis convallis." +
      "Tempus leo eu aenean sed diam urna tempor. Pulvinar vivamus fringilla";

  • #7786 33ffcd5 Thanks @daivinhtran! - Fixed #7601: Properly match Grit plugin’s code snippet with only one child.

  • #7901 0d17b05 Thanks @ematipico! - Fixed #7837, where Biome Language Server panicked when opening HTML-ish files when the experimental full support is enabled.

2.3.3

Patch Changes

  • #7907 57bd662 Thanks @ematipico! - Fixed #7839. Now the Biome parser correctly parses the Astro frontmatter even when a triple fence is inside quotes.

  • #7934 a35c496 Thanks @alissonlauffer! - Fixed #7919: The HTML parser now correctly handles Unicode BOM (Byte Order Mark) characters at the beginning of HTML files, ensuring proper parsing and tokenization.

  • #7869 c80361d Thanks @matanshavit! - Fixed #7864: Biome now preserves component tag name casing in Svelte, Astro, and Vue files.

  • #7926 69cecec Thanks @matanshavit! - Added the rule noParametersOnlyUsedInRecursion.

    This rule detects function parameters that are exclusively used in recursive calls and can be removed to simplify the function signature since they are effectively unused.

    function factorial(n, acc) {
      if (n === 0) return 1;
      return factorial(n - 1, acc); // acc is only used here
    }

    Fixes #6484.

  • #7774 2509b91 Thanks @dibashthapa! - Fixed #7657: Added the new rule no-unknown-property from ESLint

  • #7918 7165d06 Thanks @dyc3! - Fixed #7913: The CSS parser, with tailwindDirectives enabled, will now correctly handle @slot.

  • #7959 ffae203 Thanks @siketyan! - Fixed the Biome Language Server so it no longer returns an internal error when the formatter is disabled in the configuration.

2.3.4

Patch Changes

  • #7989 4855c4a Thanks @alissonlauffer! - Fixed a regression in Astro frontmatter parsing where comments inside quoted strings were incorrectly detected as actual comments. This caused the parser to prematurely terminate frontmatter parsing when encountering strings like const test = "//";. For example, the following Astro frontmatter now parses correctly:

    ---
    const test = "// not a real comment";
    ---

  • #7968 0b28f5f Thanks @denbezrukov! - Refactored formatter to use strict Token element for better performance. The new Token variant is optimized for static, ASCII-only text (keywords, operators, punctuation) with the following constraints:

    • ASCII only (no Unicode characters)
    • No newlines (\n, \r)
    • No tab characters (\t)

    This enables faster printing and fitting logic by using bulk string operations (push_str, len()) instead of character-by-character iteration with Unicode width calculations.

  • #7941 19b8280 Thanks @Conaclos! - Fixed #7943. Rules’ options are now properly merged with the inherited options from a shared configuration.

    This means that you can now override a specific option from a rule without resetting the other options to their default.

    Given the following shared configuration:

    {
      "linter": {
        "rules": {
          "style": {
            "useNamingConvention": {
              "level": "on",
              "options": {
                "strictCase": false,
                "conventions": [
                  {
                    "selector": { "kind": "variable", "scope": "global" },
                    "formats": ["CONSTANT_CASE"]
                  }
                ]
              }
            }
          }
        }
      }
    }

    And the user configuration that extends this shared configuration:

    {
      "extends": ["shared.json"],
      "linter": {
        "rules": {
          "style": {
            "useNamingConvention": {
              "level": "on",
              "options": { "strictCase": true }
            }
          }
        }
      }
    }

    The obtained merged configuration is now as follows:

    {
      "extends": ["shared.json"],
      "linter": {
        "rules": {
          "style": {
            "useNamingConvention": {
              "level": "on",
              "options": {
                "strictCase": true,
                "conventions": [
                  {
                    "selector": { "kind": "variable", "scope": "global" },
                    "formats": ["CONSTANT_CASE"]
                  }
                ]
              }
            }
          }
        }
      }
    }

  • #7969 425963d Thanks @ematipico! - Added support for the Svelte syntax {@debug}. The Biome HTML parser is now able to parse and format the blocks:

    {@debug     foo,bar,    something}
    {@debug foo, bar, something}

  • #7986 3256f82 Thanks @lisiur! - Fixed #7981. Now Biome correctly detects and parses lang='tsx' and lang='jsx' languages when used inside in .vue files, when .experimentalFullSupportEnabled is enabled.

  • #7921 547c2da Thanks @dyc3! - Fixed #7854: The CSS parser, with tailwindDirectives enabled, will now parse @source inline("underline");.

  • #7856 c9e20c3 Thanks @Netail! - Added the nursery rule noContinue. Disallowing the usage of the continue statement, structured control flow statements such as if should be used instead.

    Invalid:

    let sum = 0,
      i;
    

    for (i = 0; i < 10; i++) {
      if (i >= 5) {
        continue;
      }
    

      sum += i;
    }

    Valid:

    let sum = 0,
      i;
    

    for (i = 0; i < 10; i++) {
      if (i < 5) {
        sum += i;
      }
    }

2.3.5

Patch Changes

  • #8023 96f3e77 Thanks @ematipico! - Added support Svelte syntax {@html}. Biome now is able to parse and format the Svelte syntax {@html}:

    {@html   'div'}
    {@html 'div'}

    The contents of the expressions inside the {@html <expression>} aren’t formatted yet.

  • #8058 5f68bcc Thanks @ematipico! - Fixed a bug where the Biome Language Server would enable its project file watcher even when no project rules were enabled.

    Now the watching of nested configuration files and nested ignore files is delegated to the editor, if their LSP spec supports it.

  • #8023 96f3e77 Thanks @ematipico! - Added support Svelte syntax {@render}. Biome now is able to parse and format the Svelte syntax {@render}:

    {@render   sum(1, 2)   }
    {@render sum(1, 2)}

    The contents of the expressions inside the {@render <expression>} aren’t formatted yet.

  • #8006 f0612a5 Thanks @Bertie690! - Updated documentation and diagnostic for lint/complexity/noBannedTypes. The rule should have a more detailed description and diagnostic error message.

  • #8039 da70d8b Thanks @PFiS1737! - Biome now keeps a blank line after the frontmatter section in Astro files.

  • #8042 b7efa6f Thanks @dyc3! - The CSS Parser, with tailwindDirectives enabled, will now accept at rules like @media and @supports in @custom-variant shorthand syntax.

  • #8064 3ff9d45 Thanks @dibashthapa! - Fixed #7967: Fixed the issue with support for advanced SVG props

  • #8023 96f3e77 Thanks @ematipico! - Added support Svelte syntax {@attach}. Biome now is able to parse and format the Svelte syntax {@attach}:

    <div {@attach    myAttachment   }>...</div>
    <div {@attach myAttachment}>...</div>

    The contents of the expressions inside the {@attach <expression>} aren’t formatted yet.

  • #8001 6e8a50e Thanks @ematipico! - Added support Svelte syntax {#key}. Biome now is able to parse and format the Svelte syntax {#key}:

    {#key   expression} <div></div> {/key}
    {#key expression}
      <div></div>
    {/key}

    The contents of the expressions inside the {#key <expression>} aren’t formatted yet.

  • #8023 96f3e77 Thanks @ematipico! - Added support Svelte syntax {@const}. Biome now is able to parse and format the Svelte syntax {@const}:

    {@const   name = value}
    {@const name = value}

    The contents of the expressions inside the {@const <expression>} aren’t formatted yet.

  • #8044 8f77d4a Thanks @Netail! - Corrected rule source references. biome migrate eslint should do a bit better detecting rules in your eslint configurations.

  • #8065 1a2d1af Thanks @Netail! - Added the nursery rule useArraySortCompare. Require Array#sort and Array#toSorted calls to always provide a compareFunction.

    Invalid:

    const array = [];
    array.sort();

    Valid:

    const array = [];
    array.sort((a, b) => a - b);

  • #7673 a3a713d Thanks @dyc3! - The HTML parser is now able to parse vue directives. This enables us to write/port Vue lint rules that require inspecting the <template> section. However, this more complex parsing may result in parsing errors where there was none before. For those of you that have opted in to the experimental support (aka experimentalFullSupportEnabled), we greatly appreciate your help testing this out, and your bug reports.

  • #8031 fa6798a Thanks @ematipico! - Added support for the Svelte syntax {#if}{/if}. The Biome HTML parser is now able to parse and format the {#if}{/if} blocks:

    <!-- if / else-if / else -->
    {#if porridge.temperature > 100}
    <p>too hot!</p>
      <p>too hot!</p>
    {:else if 80 > porridge.temperature}
    <p>too cold!</p>
      <p>too cold!</p>
    {:else if 100 > porridge.temperature}
    <p>too too cold!</p>
      <p>too too cold!</p>
    {:else}
    <p>just right!</p>
      <p>just right!</p>
    {/if}

  • #8041 beeb7bb Thanks @dyc3! - The CSS parser, with tailwindDirectives enabled, will now accept lists of selectors in @custom-variant shorthand syntax.

    @custom-variant cell (th:has(&), td:has(&));

  • #8028 c09e45c Thanks @fmajestic! - The GitLab reporter now outputs format errors.

  • #8037 78011b1 Thanks @PFiS1737! - indentScriptAndStyle no longer indents the frontmatter in Astro files.

  • #8009 6374b1f Thanks @tmcw! - Fixed an edge case in the useArrowFunction rule.

    The rule no longer emits diagnostics for or offers to fix functions that reference the arguments object, because that object is undefined for arrow functions.

    Valid example:

    // Valid: this function cannot be transformed into an arrow function because
    // arguments is not defined for arrow functions.
    const getFirstArg = function () {
      return arguments[0];
    };

2.3.6 Latest

Patch Changes

  • #8100 82b9a8e Thanks @Netail! - Added the nursery rule useFind. Enforce the use of Array.prototype.find() over Array.prototype.filter() followed by [0] when looking for a single result.

    Invalid:

    [1, 2, 3].filter((x) => x > 1)[0];
    

    [1, 2, 3].filter((x) => x > 1).at(0);

  • #8118 dbc7021 Thanks @hirokiokada77! - Fixed #8117: useValidLang now accepts valid BCP 47 language tags with script subtags.

    Valid:

    <html lang="zh-Hans-CN"></html>

  • #7672 f1d5725 Thanks @Netail! - Added the nursery rule useConsistentGraphqlDescriptions, requiring all descriptions to follow the same style (either block or inline) inside GraphQL files.

    Invalid:

    enum EnumValue {
      "this is a description"
      DEFAULT
    }

    Valid:

    enum EnumValue {
      """
      this is a description
      """
      DEFAULT
    }

  • #8026 f102661 Thanks @matanshavit! - Fixed #8004: noParametersOnlyUsedInRecursion now correctly detects recursion by comparing function bindings instead of just names.

    Previously, the rule incorrectly flagged parameters when a method had the same name as an outer function but called the outer function (not itself):

    function notRecursive(arg) {
      return arg;
    }
    

    const obj = {
      notRecursive(arg) {
        return notRecursive(arg); // This calls the outer function, not the method itself
      },
    };

    Biome now properly distinguishes between these cases and will not report false positives.

  • #8097 5fc5416 Thanks @dyc3! - Added the nursery rule noVueVIfWithVFor. This rule disallows v-for and v-if on the same element.

    <!-- Invalid -->
    <div v-for="item in items" v-if="item.isActive">
      {{ item.name }}
    </div>

  • #8085 7983940 Thanks @Netail! - Added the nursery rule noForIn. Disallow iterating using a for-in loop.

    Invalid:

    for (const i in array) {
      console.log(i, array[i]);
    }

  • #8086 2b41e82 Thanks @matanshavit! - Fixed #8045: The noNestedTernary rule now correctly detects nested ternary expressions even when they are wrapped in parentheses (e.g. foo ? (bar ? 1 : 2) : 3).

    Previously, the rule would not flag nested ternaries like foo ? (bar ? 1 : 2) : 3 because the parentheses prevented detection. The rule now looks through parentheses to identify nested conditionals.

    Previously not detected (now flagged):

    const result = foo ? (bar ? 1 : 2) : 3;

    Still valid (non-nested with parentheses):

    const result = foo ? bar : baz;

  • #8075 e403868 Thanks @YTomm! - Fixed #7948: The useReadonlyClassProperties code fix when checkAllProperties is enabled will no longer insert a newline after readonly and the class property.

  • #8102 47d940e Thanks @lucasweng! - Fixed #8027. useReactFunctionComponents no longer reports class components that implement componentDidCatch using class expressions.

    The rule now correctly recognizes error boundaries defined as class expressions:

    const ErrorBoundary = class extends Component {
      componentDidCatch(error, info) {}
    

      render() {
        return this.props.children;
      }
    };

  • #8097 5fc5416 Thanks @dyc3! - Added the nursery rule useVueHyphenatedAttributes, which encourages using kebab case for attribute names, per the Vue style guide’s recommendations.

    <!-- Invalid -->
    <MyComponent myProp="value" />
    

    <!-- Valid -->
    <MyComponent my-prop="value" />

  • #8108 0f0a658 Thanks @Netail! - Added the nursery rule noSyncScripts. Prevent the usage of synchronous scripts.

    Invalid:

    <script src="https://third-party-script.js" />

    Valid:

    <script src="https://third-party-script.js" async />
    <script src="https://third-party-script.js" defer />

  • #8098 1fdcaf0 Thanks @Jayllyz! - Added documentation URLs to rule descriptions in the JSON schema.

  • #8097 5fc5416 Thanks @dyc3! - Fixed an issue with the HTML parser where it would treat Vue directives with dynamic arguments as static arguments instead.

  • #7684 f4433b3 Thanks @vladimir-ivanov! - Changed noUnusedPrivateClassMembers to align more fully with meaningful reads.

    This rule now distinguishes more carefully between writes and reads of private class members.

    • A meaningful read is any access that affects program behavior.
    • For example, this.#x += 1 both reads and writes #x, so it counts as usage.
    • Pure writes without a read (e.g. this.#x = 1 with no getter) are no longer treated as usage.

    This change ensures that private members are only considered “used” when they are actually read in a way that influences execution.

    Invalid examples (previously valid)

    class UsedMember {
      set #x(value) {
        doSomething(value);
      }
    

      foo() {
        // This assignment does not actually read #x, because there is no getter.
        // Previously, this was considered a usage, but now it’s correctly flagged.
        this.#x = 1;
      }
    }

    Valid example (Previously invalid)

    class Foo {
      #usedOnlyInWriteStatement = 5;
    

      method() {
        // This counts as a meaningful read because we both read and write the value.
        this.#usedOnlyInWriteStatement += 42;
      }
    }

  • #7684 f4433b3 Thanks @vladimir-ivanov! - Improved detection of used private class members

    The analysis for private class members has been improved: now the tool only considers a private member “used” if it is actually referenced in the code.

    • Previously, some private members might have been reported as used even if they weren’t actually accessed.
    • With this change, only members that are truly read or called in the code are counted as used.
    • Members that are never accessed will now be correctly reported as unused.

    This makes reports about unused private members more accurate and helps you clean up truly unused code.

    Example (previously valid)

    type YesNo = "yes" | "no";
    

    export class SampleYesNo {
      private yes: () => void;
      private no: () => void;
      private dontKnow: () => void; // <- will now report as unused
    

      on(action: YesNo): void {
        this[action]();
      }
    }

  • #7681 b406db6 Thanks @kedevked! - Added the new lint rule, useSpread, ported from the ESLint rule prefer-spread.

    This rule enforces the use of the spread syntax (...) over Function.prototype.apply() when calling variadic functions, as spread syntax is generally more concise and idiomatic in modern JavaScript (ES2015+).

    The rule provides a safe fix.

    Invalid

    Math.max.apply(Math, args);
    foo.apply(undefined, args);
    obj.method.apply(obj, args);

    Valid

    Math.max(...args);
    foo(...args);
    obj.method(...args);
    

    // Allowed: cases where the `this` binding is intentionally changed
    foo.apply(otherObj, args);

  • #7287 aa55c8d Thanks @ToBinio! - Fixed #7205: The noDuplicateTestHooks rule now treats chained describe variants (e.g., describe.each/for/todo) as proper describe scopes, eliminating false positives.

    The following code will no longer be a false positive:

    describe("foo", () => {
      describe.for([])("baz", () => {
        beforeEach(() => {});
      });
    

      describe.todo("qux", () => {
        beforeEach(() => {});
      });
    

      describe.todo.each([])("baz", () => {
        beforeEach(() => {});
      });
    });

  • #8013 0c0edd4 Thanks @Jayllyz! - Added the GraphQL nursery rule useUniqueGraphqlOperationName. This rule ensures that all GraphQL operations within a document have unique names.

    Invalid:

    query user {
      user {
        id
      }
    }
    

    query user {
      user {
        id
        email
      }
    }

    Valid:

    query user {
      user {
        id
      }
    }
    

    query userWithEmail {
      user {
        id
        email
      }
    }

  • #8084 c2983f9 Thanks @dyc3! - Fixed #8080: The HTML parser, when parsing Vue, can now properly handle Vue directives with no argument, modifiers, or initializer (e.g. v-else). It will no longer treat subsequent valid attributes as bogus.

    <p v-else class="flex">World</p>
    <!-- Fixed: class now gets parsed as it's own attribute -->

  • #8104 041196b Thanks @Conaclos! - Fixed noInvalidUseBeforeDeclaration. The rule no longer reports a use of an ambient variable before its declarations. The rule also completely ignores TypeScript declaration files. The following code is no longer reported as invalid:

    CONSTANT;
    declare const CONSTANT: number;

  • #8060 ba7b076 Thanks @dyc3! - Added the nursery rule useVueValidVBind, which enforces the validity of v-bind directives in Vue files.

    Invalid v-bind usages include:

    <Foo v-bind />
    <!-- Missing argument -->
    <Foo v-bind:foo />
    <!-- Missing value -->
    <Foo v-bind:foo.bar="baz" />
    <!-- Invalid modifier -->

  • #8113 fb8e3e7 Thanks @Conaclos! - Fixed noInvalidUseBeforeDeclaration. The rule now reports invalid use of classes, enums, and TypeScript’s import-equals before their declarations.

    The following code is now reported as invalid:

    new C();
    class C {}

  • #8077 0170dcb Thanks @dyc3! - Added the rule useVueValidVElseIf to enforce valid v-else-if directives in Vue templates. This rule reports invalid v-else-if directives with missing conditional expressions or when not preceded by a v-if or v-else-if directive.

  • #8077 0170dcb Thanks @dyc3! - Added the rule useVueValidVElse to enforce valid v-else directives in Vue templates. This rule reports v-else directives that are not preceded by a v-if or v-else-if directive.

  • #8077 0170dcb Thanks @dyc3! - Added the rule useVueValidVHtml to enforce valid usage of the v-html directive in Vue templates. This rule reports v-html directives with missing expressions, unexpected arguments, or unexpected modifiers.

  • #8077 0170dcb Thanks @dyc3! - Added the rule useVueValidVIf to enforce valid v-if directives in Vue templates. It disallows arguments and modifiers, and ensures a value is provided.

  • #8077 0170dcb Thanks @dyc3! - Added the rule useVueValidVOn to enforce valid v-on directives in Vue templates. This rule reports invalid v-on / shorthand @ directives with missing event names, invalid modifiers, or missing handler expressions.

Copyright (c) 2023-present Biome Developers and Contributors.