Skip to main content
Version: 2.23 (prerelease)

Linters and formatters

How to activate and use the Python linters and formatters bundled with Pants.


Benefit of Pants: consistent interface

pants lint and pants fmt will consistently and correctly run all your linters and formatters. No need to remember how to invoke each tool, and no need to write custom scripts.

This consistent interface even works with multiple languages, like running Python linters at the same time as Go, Shell, Java, and Scala.

Benefit of Pants: concurrency

Pants does several things to speed up running formatters and linters:

  • Automatically configures tools that support concurrency (e.g. a --jobs option) based on your number of cores and what else is already running.
  • Runs everything in parallel with the lint goal (although not the fmt goal, which pipes the results of one formatter to the next for correctness).
  • Runs in batches of 256 files by default, which gives parallelism even for tools that don't have a --jobs option. This also increases cache reuse.

Activating linters and formatters

Linter/formatter support is implemented in separate backends so that they are easy to opt in to individually:

BackendTool
pants.backend.python.lint.banditBandit: security linter
pants.backend.python.lint.blackBlack: code formatter
pants.backend.python.lint.docformatterDocformatter: docstring formatter
pants.backend.python.lint.flake8Flake8: style and bug linter
pants.backend.python.lint.isortisort: import statement formatter
pants.backend.python.lint.pydocstylePydocstyle: docstring linter
pants.backend.python.lint.pylintPylint: style and bug linter
pants.backend.python.lint.yapfYapf: code formatter
pants.backend.python.lint.autoflakeAutoflake: remove unused imports
pants.backend.python.lint.pyupgradePyupgrade: automatically update code to use modern Python idioms like f-strings
pants.backend.experimental.python.lint.ruff.checkRuff (for linting): an extremely fast Python linter, written in Rust.
pants.backend.experimental.python.lint.ruff.formatRuff (for formatting): an extremely fast Python code formatter, written in Rust.

To enable, add the appropriate backends in pants.toml:

pants.toml
[GLOBAL]
...
backend_packages = [
'pants.backend.python',
'pants.backend.python.lint.black',
'pants.backend.python.lint.isort',
]

You should now be able to run pants lint, and possibly pants fmt:

$ pants lint src/py/project.py
17:54:32.51 [INFO] Completed: lint - Flake8 succeeded.
17:54:32.70 [INFO] Completed: lint - Black succeeded.
All done! ✨ 🍰 ✨
1 file would be left unchanged.

17:54:33.91 [INFO] Completed: lint - isort succeeded.

✓ Black succeeded.
✓ Flake8 succeeded.
✓ isort succeeded.
How to activate MyPy

MyPy is run with the check goal, rather than lint.

Configuring the tools, for example, adding plugins

You can configure each formatter and linter using these options:

OptionWhat it does
interpreter_constraintsWhat interpreter to run the tool with. (bandit, flake8, and pylint determine this based on your code's interpreter constraints.)
argsAny command-line arguments you want to pass to the tool.
configPath to a config file. Useful if the file is in a non-standard location such that it cannot be auto-discovered.
install_from_resolveName of a custom resolve to use for tool versions and plugins. See Lockfiles for tools.
requirementsOptional subset of requirements to install from the custom resolve for this formatter or linter. See Lockfiles for tools.

For example:

pants.toml
[docformatter]
args = ["--wrap-summaries=100", "--wrap-descriptions=100"]

[python.resolves]
# A custom resolve that updates the version and adds a custom plugin.
flake8 = "3rdparty/python/flake8.lock"

[flake8]
# Load a config file in a non-standard location.
config = "build-support/flake8"
install_from_resolve = "flake8"

Then set up the resolve's inputs:

python_requirements(
source="flake8-requirements.txt",
resolve="flake8",
)

And generate its custom lockfile:

Bash
$ pants generate-lockfiles --resolve=flake8
16:00:39.26 [INFO] Completed: Generate lockfile for flake8
16:00:39.29 [INFO] Wrote lockfile for the resolve `flake8` to 3rdparty/python/flake8.lock

Run pants help-advanced black, pants help-advanced flake8, and so on for more information.

Config files are normally auto-discovered

For tools that autodiscover config files—such as Black, isort, Flake8, and Pylint—Pants will include any relevant config files in the process's sandbox when running the tool.

If your config file is in a non-standard location, you must instead set the --config option, e.g. [isort].config. This will ensure that the config file is included in the process's sandbox and Pants will instruct the tool to load the config.

Running only certain formatters or linters

To temporarily skip a tool, use the --skip option for that tool. For example, run:

❯  pants --black-skip --flake8-skip lint ::

You can also use the --lint-only and --fmt-only options with the names of the tools:

❯ pants lint --only=black ::

# To run several, you can use either approach:
❯ pants fmt --only=black --only=isort ::
❯ pants fmt --only='["black", "isort"]' ::

You can also skip for certain targets with the skip_<tool> fields, which can be useful for incrementally adopting new tools. For example:

project/BUILD
python_sources(
name="lib",
# Skip Black for all non-test files in this folder.
skip_black=True,
overrides={
"strutil.py": {"skip_flake8": True},
("docutil.py", "dirutil.py"): {"skip_isort": True},
},
)

python_tests(
name="tests",
# Skip isort for all the test files in this folder.
skip_isort=True,
)

When you run pants fmt and pants lint, Pants will ignore any files belonging to skipped targets.

Tip: only run over changed files

With formatters and linters, there is usually no need to rerun on files that have not changed.

Use the option --changed-since to get much better performance, like this:

❯ pants --changed-since=HEAD fmt

or

❯ pants --changed-since=main lint

Pants will find which files have changed and only run over those files. See Advanced target selection for more information.

Tips for specific tools

Order of backend_packages matters for fmt

Pants will run formatters in the order in which they appear in the backend_packages option.

For example, you likely want to put Autoflake (which removes unused imports) before Black and Isort, which will format your import statements.

pants.toml
[GLOBAL]
backend_packages = [
# Note that we want Autoflake to run before Black and isort,
# so it must appear first.
"pants.backend.python.experimental.autoflake",
"pants.backend.python.black",
"pants.backend.python.isort",
]

Bandit, Flake8, and Pylint: report files

Flake8, Bandit, and Pylint can generate report files saved to disk.

For Pants to properly preserve the reports, instruct the tools to write to the reports/ folder by updating their config files, or --flake8-args, --bandit-args, and --pylint-args. For example, in your pants.toml:

[bandit]
args = ["--output=reports/report.txt"]

[flake8]
args = ["--output-file=reports/report.txt"]

[pylint]
args = ["--output-format=text:reports/report.txt"]

Pants will copy all reports into the folder dist/lint/<linter_name>.

Pylint and Flake8: how to add first-party plugins

See [pylint].source_plugins and [flake8].source_plugins for instructions to add plugins written by you.

Bandit: less verbose logging

Bandit output can be extremely verbose, including on successful runs. You may want to use its --quiet option, which will turn off output for successful runs but keep it for failures.

For example, you can set this in your pants.toml:

[bandit]
args = ["--quiet"]

Black and isort can work together

If you use both black and isort, you most likely will need to tell isort to work in a mode compatible with black. It is also a good idea to ensure they use the same line length. This requires tool specific configuration, which could go into pyproject.toml for example:

# pyproject.toml
[tool.isort]
profile = "black"
line_length = 100

[tool.black]
line-length = 100

Pyupgrade: specify which Python version to target

You must tell Pyupgrade which version of Python to target, like this:

# pants.toml
[pyupgrade]
args = ["--py36-plus"]

isort: possible issues with its import classifier algorithm

Some Pants users had to explicitly set default_section = "THIRDPARTY" to get iSort 5 to correctly classify their first-party imports, even though this is the default value.

They report that this config works for them:

# pyproject.toml
[tool.isort]
known_first_party = ["my_org"]
default_section = "THIRDPARTY"

You may also want to try downgrading to iSort 4.x by setting version = "isort>=4.6,<5" in the [isort] options scope.

Black and isort: excluding files

Although it is possible to skip linters for individual targets using skip_ parameters, sometimes an entire category of files needs to be excluded, e.g. generated files with a prefix or suffix. This can be done by configuring those tools directly within pyproject.toml.

In order to exclude files from being formatted by Black and isort, it may be necessary to tell those tools to respect skip configuration options. Since Pants runs these tools on individual files rather than directories, they often run on every Python file in spite of the intuitive skip arguments. See e.g. this post from the Black maintainers.

Pants users report that this config works for them:

# pyproject.toml
[tool.isort]
# tell isort to respect skip_glob
filter_files = true
# in particular, extend_skip_glob doesn't seem to work under Pants isort
skip_glob = [
"**/*_skip_me.py",
"**/*_skip_me.pyi",
]

[tool.black]
# in particular, extend-exclude and exclude will not work
force-exclude='''
^(
.*_skip_me\.py(i)?
)$
'''