An extremely fast Python linter, written in Rust.

Linting the CPython codebase from scratch.

• ⚡️ 10-100x faster than existing linters
• 🐍 Installable via pip
• 🤝 Python 3.10 compatibility
• 🛠️ pyproject.toml support
• 📦 ESLint-inspired cache support
• 🔧 ESLint-inspired autofix support (e.g., automatically remove unused imports)
• 👀 TypeScript-inspired --watch support, for continuous file monitoring
• ⚖️ Near-parity with the built-in Flake8 rule set
• 🔌 Native re-implementations of popular Flake8 plugins, like flake8-docstrings (pydocstyle)

Ruff aims to be orders of magnitude faster than alternative tools while integrating more functionality behind a single, common interface. Ruff can be used to replace Flake8 (plus a variety of plugins), pydocstyle, yesqa, and even a subset of pyupgrade and autoflake all while executing tens or hundreds of times faster than any individual tool.

Read the launch blog post.

1. Installation and Usage
2. Configuration
3. Supported Rules
4. Editor Integrations
5. FAQ
6. Development
7. Releases
8. Benchmarks
9. License
10. Contributing

Available as ruff on PyPI:

pip install ruff

To run Ruff, try any of the following:

ruff path/to/code/to/check.py
ruff path/to/code/
ruff path/to/code/*.py

You can run Ruff in --watch mode to automatically re-run on-change:

ruff path/to/code/ --watch

Ruff also works with pre-commit:

repos:
  - repo: https://github.com/charliermarsh/ruff-pre-commit
    rev: v0.0.82
    hooks:
      - id: lint

Ruff is configurable both via pyproject.toml and the command line.

For example, you could configure Ruff to only enforce a subset of rules with:

[tool.ruff]
line-length = 88
select = [
    "F401",
    "F403",
]

Alternatively, on the command-line:

ruff path/to/code/ --select F401 --select F403

See ruff --help for more:

ruff: An extremely fast Python linter.

Usage: ruff [OPTIONS] <FILES>...

Arguments:
  <FILES>...

Options:
      --config <CONFIG>
          Path to the `pyproject.toml` file to use for configuration
  -v, --verbose
          Enable verbose logging
  -q, --quiet
          Disable all logging (but still exit with status code "1" upon detecting errors)
  -e, --exit-zero
          Exit with status code "0", even upon detecting errors
  -w, --watch
          Run in watch mode by re-running whenever files change
  -f, --fix
          Attempt to automatically fix lint errors
  -n, --no-cache
          Disable cache reads
      --select <SELECT>
          List of error codes to enable
      --extend-select <EXTEND_SELECT>
          Like --select, but adds additional error codes on top of the selected ones
      --ignore <IGNORE>
          List of error codes to ignore
      --extend-ignore <EXTEND_IGNORE>
          Like --ignore, but adds additional error codes on top of the ignored ones
      --exclude <EXCLUDE>
          List of paths, used to exclude files and/or directories from checks
      --extend-exclude <EXTEND_EXCLUDE>
          Like --exclude, but adds additional files and directories on top of the excluded ones
      --per-file-ignores <PER_FILE_IGNORES>
          List of mappings from file pattern to code to exclude
      --format <FORMAT>
          Output serialization format for error messages [default: text] [possible values: text, json]
      --show-files
          See the files ruff will be run against with the current settings
      --show-settings
          See ruff's settings
      --add-noqa
          Enable automatic additions of noqa directives to failing lines
      --dummy-variable-rgx <DUMMY_VARIABLE_RGX>
          Regular expression matching the name of dummy variables
      --target-version <TARGET_VERSION>
          The minimum Python version that should be supported
      --stdin-filename <STDIN_FILENAME>
          The name of the file when passing it through stdin
  -h, --help
          Print help information
  -V, --version
          Print version information

Exclusions are based on globs, and can be either:

• Single-path patterns, like .mypy_cache (to exclude any directory named .mypy_cache in the tree), foo.py (to exclude any file named foo.py), or foo_*.py (to exclude any file matching foo_*.py ).
• Relative patterns, like directory/foo.py (to exclude that specific file) or directory/*.py (to exclude any Python files in directory). Note that these paths are relative to the project root (e.g., the directory containing your pyproject.toml).

To omit a lint check entirely, add it to the "ignore" list via --ignore or --extend-ignore, either on the command-line or in your project.toml file.

To ignore an error in-line, Ruff uses a noqa system similar to Flake8. To ignore an individual error, add # noqa: {code} to the end of the line, like so:

# Ignore F841.
x = 1  # noqa: F841

# Ignore E741 and F841.
i = 1  # noqa: E741, F841

# Ignore _all_ errors.
x = 1  # noqa

Note that, for multi-line strings, the noqa directive should come at the end of the string, and will apply to the entire body, like so:

"""Lorem ipsum dolor sit amet.

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
"""  # noqa: E501

Ruff supports several workflows to aid in noqa management.

First, Ruff provides a special error code, M001, to enforce that your noqa directives are "valid", in that the errors they say they ignore are actually being triggered on that line (and thus suppressed). You can run ruff /path/to/file.py --extend-select M001 to flag unused noqa directives.

Second, Ruff can automatically remove unused noqa directives via its autofix functionality. You can run ruff /path/to/file.py --extend-select M001 --fix to automatically remove unused noqa directives.

Third, Ruff can automatically add noqa directives to all failing lines. This is useful when migrating a new codebase to Ruff. You can run ruff /path/to/file.py --add-noqa to automatically add noqa directives to all failing lines, with the appropriate error codes.

By default, Ruff enables all E, W, and F error codes, which correspond to those built-in to Flake8.

The 🛠 emoji indicates that a rule is automatically fixable by the --fix command-line option.

Ruff can be installed as an External Tool in PyCharm. Open the Preferences pane, then navigate to "Tools", then "External Tools". From there, add a new tool with the following configuration:

Ruff should then appear as a runnable action:

GitHub Actions has everything you need to run Ruff out-of-the-box:

name: CI
on: push
jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - name: Install Python
        uses: actions/setup-python@v4
        with:
          python-version: "3.10"
      - name: Install dependencies
        run: |
          python -m pip install --upgrade pip
          pip install ruff
      - name: Run Ruff
        run: ruff .

Yes. Ruff is compatible with Black out-of-the-box, as long as the line-length setting is consistent between the two.

As a project, Ruff is designed to be used alongside Black and, as such, will defer implementing stylistic lint rules that are obviated by autoformatting.

Ruff can be used as a (near) drop-in replacement for Flake8 when used (1) without or with a small number of plugins, (2) alongside Black, and (3) on Python 3 code.

Under those conditions Ruff is missing 14 rules related to string .format calls, 1 rule related to docstring parsing, and 1 rule related to redefined variables.

Ruff re-implements some of the most popular Flake8 plugins and related code quality tools natively, including:

• pydocstyle
• yesqa
• flake8-docstrings
• flake8-builtins
• flake8-super
• flake8-print
• flake8-comprehensions
• flake8-bugbear (3/32)
• pyupgrade (8/34)
• autoflake (1/7)

Beyond rule-set parity, Ruff suffers from the following limitations vis-à-vis Flake8:

1. Ruff does not yet support a few Python 3.9 and 3.10 language features, including structural pattern matching and parenthesized context managers.
2. Flake8 has a plugin architecture and supports writing custom lint rules. (To date, popular Flake8 plugins have been re-implemented within Ruff directly.)

Today, Ruff can be used to replace Flake8 when used with any of the following plugins:

• flake8-docstrings
• flake8-builtins
• flake8-super
• flake8-print
• flake8-comprehensions
• flake8-bugbear (3/32)

Ruff also implements the functionality that you get from yesqa, and a subset of the rules implemented in pyupgrade (8/34).

If you're looking to use Ruff, but rely on an unsupported Flake8 plugin, free to file an Issue.

Nope! Ruff is available as ruff on PyPI:

pip install ruff

Ruff ships with wheels for all major platforms, which enables pip to install Ruff without relying on Rust at all.

Ruff does not yet support third-party plugins, though a plugin system is within-scope for the project. See #283 for more.

Yes! To enable a specific docstring convention, start by enabling all pydocstyle error codes, and then selectively disabling based on your preferred convention.

For example, if you're coming from flake8-docstrings, the following configuration is equivalent to --docstring-convention numpy:

[tool.ruff]
extend-select = [
    "D100",
    "D101",
    "D102",
    "D103",
    "D104",
    "D105",
    "D106",
    "D200",
    "D201",
    "D202",
    "D204",
    "D205",
    "D206",
    "D207",
    "D208",
    "D209",
    "D210",
    "D211",
    "D214",
    "D215",
    "D300",
    "D400",
    "D402",
    "D403",
    "D404",
    "D405",
    "D406",
    "D407",
    "D408",
    "D409",
    "D410",
    "D411",
    "D412",
    "D413",
    "D418",
    "D419",
]

Similarly, the following is equivalent to --docstring-convention google:

[tool.ruff]
extend-select = [
    "D100",
    "D101",
    "D102",
    "D103",
    "D104",
    "D105",
    "D106",
    "D107",
    "D200",
    "D201",
    "D202",
    "D205",
    "D206",
    "D207",
    "D208",
    "D209",
    "D210",
    "D211",
    "D212",
    "D214",
    "D300",
    "D402",
    "D403",
    "D405",
    "D410",
    "D411",
    "D412",
    "D414",
    "D415",
    "D416",
    "D417",
    "D418",
    "D419",
]

Ruff is written in Rust (1.63.0). You'll need to install the Rust toolchain for development.

Assuming you have cargo installed, you can run:

cargo run resources/test/fixtures
cargo fmt
cargo clippy
cargo test

Ruff is distributed on PyPI, and published via maturin.

See: .github/workflows/release.yaml.

First, clone CPython. It's a large and diverse Python codebase, which makes it a good target for benchmarking.

git clone --branch 3.10 https://github.com/python/cpython.git resources/test/cpython

Add this pyproject.toml to the CPython directory:

[tool.ruff]
line-length = 88
extend-exclude = [
    "Lib/lib2to3/tests/data/bom.py",
    "Lib/lib2to3/tests/data/crlf.py",
    "Lib/lib2to3/tests/data/different_encoding.py",
    "Lib/lib2to3/tests/data/false_encoding.py",
    "Lib/lib2to3/tests/data/py2_test_grammar.py",
    "Lib/test/bad_coding2.py",
    "Lib/test/badsyntax_3131.py",
    "Lib/test/badsyntax_pep3120.py",
    "Lib/test/encoded_modules/module_iso_8859_1.py",
    "Lib/test/encoded_modules/module_koi8_r.py",
    "Lib/test/test_fstring.py",
    "Lib/test/test_grammar.py",
    "Lib/test/test_importlib/test_util.py",
    "Lib/test/test_named_expressions.py",
    "Lib/test/test_patma.py",
    "Lib/test/test_source_encoding.py",
    "Tools/c-analyzer/c_parser/parser/_delim.py",
    "Tools/i18n/pygettext.py",
    "Tools/test2to3/maintest.py",
    "Tools/test2to3/setup.py",
    "Tools/test2to3/test/test_foo.py",
    "Tools/test2to3/test2to3/hello.py",
]

Next, to benchmark the release build:

cargo build --release

hyperfine --ignore-failure --warmup 10 --runs 100 \
  "./target/release/ruff ./resources/test/cpython/ --no-cache" \
  "./target/release/ruff ./resources/test/cpython/"

Benchmark 1: ./target/release/ruff ./resources/test/cpython/ --no-cache
  Time (mean ± σ):     297.4 ms ±   4.9 ms    [User: 2460.0 ms, System: 67.2 ms]
  Range (min … max):   287.7 ms … 312.1 ms    100 runs

  Warning: Ignoring non-zero exit code.

Benchmark 2: ./target/release/ruff ./resources/test/cpython/
  Time (mean ± σ):      79.6 ms ±   7.3 ms    [User: 59.7 ms, System: 356.1 ms]
  Range (min … max):    62.4 ms … 111.2 ms    100 runs

  Warning: Ignoring non-zero exit code.

To benchmark against the ecosystem's existing tools:

hyperfine --ignore-failure --warmup 5 \
  "./target/release/ruff ./resources/test/cpython/ --no-cache" \
  "pylint --recursive=y resources/test/cpython/" \
  "pyflakes resources/test/cpython" \
  "autoflake --recursive --expand-star-imports --remove-all-unused-imports --remove-unused-variables --remove-duplicate-keys resources/test/cpython" \
  "pycodestyle resources/test/cpython" \
  "flake8 resources/test/cpython" \
  "python -m scripts.run_flake8 resources/test/cpython"

In order, these evaluate:

• Ruff
• Pylint
• Pyflakes
• autoflake
• pycodestyle
• Flake8
• Flake8, with a hack to enable multiprocessing on macOS

(You can poetry install from ./scripts to create a working environment for the above.)

Benchmark 1: ./target/release/ruff ./resources/test/cpython/ --no-cache
  Time (mean ± σ):     297.9 ms ±   7.0 ms    [User: 2436.6 ms, System: 65.9 ms]
  Range (min … max):   289.9 ms … 314.6 ms    10 runs

  Warning: Ignoring non-zero exit code.

Benchmark 2: pylint --recursive=y resources/test/cpython/
  Time (mean ± σ):     37.634 s ±  0.225 s    [User: 36.728 s, System: 0.853 s]
  Range (min … max):   37.201 s … 38.106 s    10 runs

  Warning: Ignoring non-zero exit code.

Benchmark 3: pyflakes resources/test/cpython
  Time (mean ± σ):     40.950 s ±  0.449 s    [User: 40.688 s, System: 0.229 s]
  Range (min … max):   40.348 s … 41.671 s    10 runs

  Warning: Ignoring non-zero exit code.

Benchmark 4: autoflake --recursive --expand-star-imports --remove-all-unused-imports --remove-unused-variables --remove-duplicate-keys resources/test/cpython
  Time (mean ± σ):     11.562 s ±  0.160 s    [User: 107.022 s, System: 1.143 s]
  Range (min … max):   11.417 s … 11.917 s    10 runs

Benchmark 5: pycodestyle resources/test/cpython
  Time (mean ± σ):     67.428 s ±  0.985 s    [User: 67.199 s, System: 0.203 s]
  Range (min … max):   65.313 s … 68.496 s    10 runs

  Warning: Ignoring non-zero exit code.

Benchmark 6: flake8 resources/test/cpython
  Time (mean ± σ):     116.099 s ±  1.178 s    [User: 115.217 s, System: 0.845 s]
  Range (min … max):   114.180 s … 117.724 s    10 runs

  Warning: Ignoring non-zero exit code.

Benchmark 7: python -m scripts.run_flake8 resources/test/cpython
  Time (mean ± σ):     20.477 s ±  0.349 s    [User: 142.372 s, System: 1.504 s]
  Range (min … max):   20.107 s … 21.183 s    10 runs

Summary
  './target/release/ruff ./resources/test/cpython/ --no-cache' ran
   38.81 ± 1.05 times faster than 'autoflake --recursive --expand-star-imports --remove-all-unused-imports --remove-unused-variables --remove-duplicate-keys resources/test/cpython'
   68.74 ± 1.99 times faster than 'python -m scripts.run_flake8 resources/test/cpython'
  126.33 ± 3.05 times faster than 'pylint --recursive=y resources/test/cpython/'
  137.46 ± 3.55 times faster than 'pyflakes resources/test/cpython'
  226.35 ± 6.23 times faster than 'pycodestyle resources/test/cpython'
  389.73 ± 9.92 times faster than 'flake8 resources/test/cpython'

MIT

Contributions are welcome and hugely appreciated. To get started, check out the contributing guidelines.