-
Default to Python 3.9
Ruff now defaults to Python 3.9 instead of 3.8 if no explicit Python version is configured using
ruff.target-version
orproject.requires-python
(#13896) -
Changed location of
pydoclint
diagnosticspydoclint
diagnostics now point to the first-line of the problematic docstring. Previously, this was not the case.If you've opted into these preview rules but have them suppressed using
noqa
comments in some places, this change may mean that you need to move thenoqa
suppression comments. Most users should be unaffected by this change. -
Use XDG (i.e.
~/.local/bin
) instead of the Cargo home directory in the standalone installerPreviously, Ruff's installer used
$CARGO_HOME
or~/.cargo/bin
for its target install directory. Now, Ruff will be installed into$XDG_BIN_HOME
,$XDG_DATA_HOME/../bin
, or~/.local/bin
(in that order).This change is only relevant to users of the standalone Ruff installer (using the shell or PowerShell script). If you installed Ruff using uv or pip, you should be unaffected.
-
Changes to the line width calculation
Ruff now uses a new version of the unicode-width Rust crate to calculate the line width. In very rare cases, this may lead to lines containing Unicode characters being reformatted, or being considered too long when they were not before (
E501
).
- The pytest rules
PT001
andPT023
now default to omitting the decorator parentheses when there are no arguments (#12838, #13292). This was a change that we attempted to make in Ruff v0.6.0, but only partially made due to an error on our part. See the blog post for more details. - The
useless-try-except
rule (in ourtryceratops
category) has been recoded fromTRY302
toTRY203
(#13502). This ensures Ruff's code is consistent with the same rule in thetryceratops
linter. - The
lint.allow-unused-imports
setting has been removed (#13677). Uselint.pyflakes.allow-unused-imports
instead.
-
Detect imports in
src
layouts by default forisort
rules (#12848) -
The pytest rules
PT001
andPT023
now default to omitting the decorator parentheses when there are no arguments (#12838). -
Lint and format Jupyter Notebook by default (#12878).
You can disable specific rules for notebooks using
per-file-ignores
:[tool.ruff.lint.per-file-ignores] "*.ipynb" = ["E501"] # disable line-too-long in notebooks
If you'd prefer to either only lint or only format Jupyter Notebook files, you can use the section-specific
exclude
option to do so. For example, the following would only lint Jupyter Notebook files and not format them:[tool.ruff.format] exclude = ["*.ipynb"]
And, conversely, the following would only format Jupyter Notebook files and not lint them:
[tool.ruff.lint] exclude = ["*.ipynb"]
You can completely disable Jupyter Notebook support by updating the
extend-exclude
setting:[tool.ruff] extend-exclude = ["*.ipynb"]
- Follow the XDG specification to discover user-level configurations on macOS (same as on other Unix platforms)
- Selecting
ALL
now excludes deprecated rules - The released archives now include an extra level of nesting, which can be removed with
--strip-components=1
when untarring. - The release artifact's file name no longer includes the version tag. This enables users to install via
/latest
URLs on GitHub.
The formatter now formats code according to the Ruff 2024.2 style guide. Read the changelog for a detailed list of stabilized style changes.
isort
: Use one blank line after imports in typing stub files (#9971)
Previously, Ruff used one or two blank lines (or the number configured by isort.lines-after-imports
) after imports in typing stub files (.pyi
files).
The typing style guide for stubs recommends using at most 1 blank line for grouping.
As of this release, isort
now always uses one blank line after imports in stub files, the same as the formatter.
build
is no longer excluded by default (#10093)
Ruff maintains a list of directories and files that are excluded by default. This list now consists of the following patterns:
.bzr
.direnv
.eggs
.git
.git-rewrite
.hg
.ipynb_checkpoints
.mypy_cache
.nox
.pants.d
.pyenv
.pytest_cache
.pytype
.ruff_cache
.svn
.tox
.venv
.vscode
__pypackages__
_build
buck-out
dist
node_modules
site-packages
venv
Previously, the build
directory was included in this list. However, the build
directory tends to be a not-unpopular directory
name, and excluding it by default caused confusion. Ruff now no longer excludes build
except if it is excluded by a .gitignore
file
or because it is listed in extend-exclude
.
Previously, ruff rule
and ruff linter
accepted the --format <FORMAT>
option as an alias for --output-format
. Ruff no longer
supports this alias. Please use ruff rule --output-format <FORMAT>
and ruff linter --output-format <FORMAT>
instead.
site-packages
is now excluded by default (#5513)
Ruff maintains a list of default exclusions, which now consists of the following patterns:
.bzr
.direnv
.eggs
.git-rewrite
.git
.hg
.ipynb_checkpoints
.mypy_cache
.nox
.pants.d
.pyenv
.pytest_cache
.pytype
.ruff_cache
.svn
.tox
.venv
.vscode
__pypackages__
_build
buck-out
build
dist
node_modules
site-packages
venv
Previously, the site-packages
directory was not excluded by default. While site-packages
tends
to be excluded anyway by virtue of the .venv
exclusion, this may not be the case when using Ruff
from VS Code outside a virtual environment.
Ruff previously used the format
setting, --format
CLI option, and RUFF_FORMAT
environment variable to
configure the output format of the CLI. This usage was deprecated in v0.0.291
— the format
setting is now used
to control Ruff's code formatting. As of this release:
- The
format
setting cannot be used to configure the output format, useoutput-format
instead - The
RUFF_FORMAT
environment variable is ignored, useRUFF_OUTPUT_FORMAT
instead - The
--format
option has been removed fromruff check
, use--output-format
instead
Unsafe fixes are not applied by default (#7769)
Ruff labels fixes as "safe" and "unsafe". The meaning and intent of your code will be retained when applying safe
fixes, but the meaning could be changed when applying unsafe fixes. Previously, unsafe fixes were always displayed
and applied when fixing was enabled. Now, unsafe fixes are hidden by default and not applied. The --unsafe-fixes
flag or unsafe-fixes
configuration option can be used to enable unsafe fixes.
See the docs for details.
Remove formatter-conflicting rules from the default rule set (#7900)
Previously, Ruff enabled all implemented rules in Pycodestyle (E
) by default. Ruff now only includes the
Pycodestyle prefixes E4
, E7
, and E9
to exclude rules that conflict with automatic formatters. Consequently,
the stable rule set no longer includes line-too-long
(E501
) and mixed-spaces-and-tabs
(E101
). Other
excluded Pycodestyle rules include whitespace enforcement in E1
and E2
; these rules are currently in preview, and are already omitted by default.
This change only affects those using Ruff under its default rule set. Users that include E
in their select
will experience no change in behavior.
Remove support for emoji identifiers (#7212)
Previously, Ruff supported the non-standard compliant emoji identifiers e.g. 📦 = 1
.
We decided to remove this non-standard language extension, and Ruff now reports syntax errors for emoji identifiers in your code, the same as CPython.
Improved GitLab fingerprints (#7203)
GitLab uses fingerprints to identify new, existing, or fixed violations. Previously, Ruff included the violation's position in the fingerprint. Using the location has the downside that changing any code before the violation causes the fingerprint to change, resulting in GitLab reporting one fixed and one new violation even though it is a pre-existing violation.
Ruff now uses a more stable location-agnostic fingerprint to minimize that existing violations incorrectly get marked as fixed and re-reported as new violations.
Expect GitLab to report each pre-existing violation in your project as fixed and a new violation in your Ruff upgrade PR.
The target Python version now defaults to 3.8 instead of 3.10 (#6397)
Previously, when a target Python version was not specified, Ruff would use a default of Python 3.10. However, it is safer to default to an older Python version to avoid assuming the availability of new features. We now default to the oldest supported Python version which is currently Python 3.8.
(We still support Python 3.7 but since it has reached EOL we've decided not to make it the default here.)
Note this change was announced in 0.0.283 but not active until 0.0.284.
.ipynb_checkpoints
, .pyenv
, .pytest_cache
, and .vscode
are now excluded by default (#5513)
Ruff maintains a list of default exclusions, which now consists of the following patterns:
.bzr
.direnv
.eggs
.git
.git-rewrite
.hg
.ipynb_checkpoints
.mypy_cache
.nox
.pants.d
.pyenv
.pytest_cache
.pytype
.ruff_cache
.svn
.tox
.venv
.vscode
__pypackages__
_build
buck-out
build
dist
node_modules
venv
Previously, the .ipynb_checkpoints
, .pyenv
, .pytest_cache
, and .vscode
directories were not
excluded by default. This change brings Ruff's default exclusions in line with other tools like
Black.
The keep-runtime-typing
setting has been reinstated (#5470)
The keep-runtime-typing
setting has been reinstated with revised semantics. This setting was
removed in #4427, as it was equivalent to ignoring
the UP006
and UP007
rules via Ruff's standard ignore
mechanism.
Taking UP006
(rewrite List[int]
to list[int]
) as an example, the setting now behaves as
follows:
- On Python 3.7 and Python 3.8, setting
keep-runtime-typing = true
will cause Ruff to ignoreUP006
violations, even iffrom __future__ import annotations
is present in the file. While such annotations are valid in Python 3.7 and Python 3.8 when combined withfrom __future__ import annotations
, they aren't supported by libraries like Pydantic and FastAPI, which rely on runtime type checking. - On Python 3.9 and above, the setting has no effect, as
list[int]
is a valid type annotation, and libraries like Pydantic and FastAPI support it without issue.
In short: keep-runtime-typing
can be used to ensure that Ruff doesn't introduce type annotations
that are not supported at runtime by the current Python version, which are unsupported by libraries
like Pydantic and FastAPI.
Note that this is not a breaking change, but is included here to complement the previous removal
of keep-runtime-typing
.
The keep-runtime-typing
setting has been removed (#4427)
Enabling the keep-runtime-typing
option, located under the pyupgrade
section, is equivalent
to ignoring the UP006
and UP007
rules via Ruff's standard ignore
mechanism. As there's no
need for a dedicated setting to disable these rules, the keep-runtime-typing
option has been
removed.
update-check
is no longer a valid configuration option (#4313)
The update-check
functionality was deprecated in #2530,
in that the behavior itself was removed, and Ruff was changed to warn when that option was enabled.
Now, Ruff will throw an error when update-check
is provided via a configuration file (e.g.,
update-check = false
) or through the command-line, since it has no effect. Users should remove
this option from their configuration.
--fix-only
now exits with a zero exit code, unless --exit-non-zero-on-fix
is specified (#4146)
Previously, --fix-only
would exit with a non-zero exit code if any fixes were applied. This
behavior was inconsistent with --fix
, and further, meant that --exit-non-zero-on-fix
was
effectively ignored when --fix-only
was specified.
Now, --fix-only
will exit with a zero exit code, unless --exit-non-zero-on-fix
is specified,
in which case it will exit with a non-zero exit code if any fixes were applied.
Fixes are now represented as a list of edits (#3709)
Previously, Ruff represented each fix as a single edit, which prohibited Ruff from automatically fixing violations that required multiple edits across a file. As such, Ruff now represents each fix as a list of edits.
This primarily affects the JSON API. Ruff's JSON representation used to represent the fix
field as
a single edit, like so:
{
"message": "Remove unused import: `sys`",
"content": "",
"location": {"row": 1, "column": 0},
"end_location": {"row": 2, "column": 0}
}
The updated representation instead includes a list of edits:
{
"message": "Remove unused import: `sys`",
"edits": [
{
"content": "",
"location": {"row": 1, "column": 0},
"end_location": {"row": 2, "column": 0},
}
]
}
multiple-statements-on-one-line-def
(E704
) was removed (#2773)
This rule was introduced in v0.0.245. However, it turns out that pycodestyle and Flake8 ignore this rule by default, as it is not part of PEP 8. As such, we've removed it from Ruff.
Ruff's public check
method was removed (#2709)
Previously, Ruff exposed a check
method as a public Rust API. This method was used by few,
if any clients, and was not well documented or supported. As such, it has been removed, with
the intention of adding a stable public API in the future.
select
, extend-select
, ignore
, and extend-ignore
have new semantics (#2312)
Previously, the interplay between select
and its related options could lead to unexpected
behavior. For example, ruff --select E501 --ignore ALL
and ruff --select E501 --extend-ignore ALL
behaved differently. (See #2312 for more
examples.)
When Ruff determines the enabled rule set, it has to reconcile select
and ignore
from a variety
of sources, including the current pyproject.toml
, any inherited pyproject.toml
files, and the
CLI.
The new semantics are such that Ruff uses the "highest-priority" select
as the basis for the rule
set, and then applies any extend-select
, ignore
, and extend-ignore
adjustments. CLI options
are given higher priority than pyproject.toml
options, and the current pyproject.toml
file is
given higher priority than any inherited pyproject.toml
files.
extend-select
and extend-ignore
are no longer given "top priority"; instead, they merely append
to the select
and ignore
lists, as in Flake8.
This change is largely backwards compatible -- most users should experience no change in behavior. However, as an example of a breaking change, consider the following:
[tool.ruff]
ignore = ["F401"]
Running ruff --select F
would previously have enabled all F
rules, apart from F401
. Now, it
will enable all F
rules, including F401
, as the command line's --select
resets the resolution.
remove-six-compat
(UP016
) has been removed (#2332)
The remove-six-compat
rule has been removed. This rule was only useful for one-time Python 2-to-3
upgrades.
--explain
, --clean
, and --generate-shell-completion
are now subcommands (#2190)
--explain
, --clean
, and --generate-shell-completion
are now implemented as subcommands:
ruff . # Still works! And will always work.
ruff check . # New! Also works.
ruff --explain E402 # Still works.
ruff rule E402 # New! Also works. (And preferred.)
# Oops! The command has to come first.
ruff --format json --explain E402 # No longer works.
ruff --explain E402 --format json # Still works!
ruff rule E402 --format json # Works! (And preferred.)
This change is largely backwards compatible -- most users should experience no change in behavior. However, please note the following exceptions:
-
Subcommands will now fail when invoked with unsupported arguments, instead of silently ignoring them. For example, the following will now fail:
ruff --clean --respect-gitignore
(the
clean
command doesn't support--respect-gitignore
.) -
The semantics of
ruff <arg>
have changed slightly when<arg>
is a valid subcommand. For example, prior to this release, runningruff rule
would runruff
over a file or directory calledrule
. Now,ruff rule
would invoke therule
subcommand. This should only impact projects with files or directories namedrule
,check
,explain
,clean
, orgenerate-shell-completion
. -
Scripts that invoke ruff should supply
--
before any positional arguments. (The semantics ofruff -- <arg>
have not changed.) -
--explain
previously treated--format grouped
as a synonym for--format text
. This is no longer supported; instead, use--format text
.
misplaced-comparison-constant
(PLC2201
) was deprecated in favor of SIM300
(#1980)
These two rules contain (nearly) identical logic. To deduplicate the rule set, we've upgraded
SIM300
to handle a few more cases, and deprecated PLC2201
in favor of SIM300
.
@functools.cache
rewrites have been moved to a standalone rule (UP033
) (#1938)
Previously, UP011
handled both @functools.lru_cache()
-to-@functools.lru_cache
conversions,
and @functools.lru_cache(maxsize=None)
-to-@functools.cache
conversions. The latter has been
moved out to its own rule (UP033
). As such, some # noqa: UP011
comments may need to be updated
to reflect the change in rule code.
--max-complexity
has been removed from the CLI (#1877)
The McCabe plugin's --max-complexity
setting has been removed from the CLI, for consistency with
the treatment of other, similar settings.
To set the maximum complexity, use the max-complexity
property in your pyproject.toml
file,
like so:
[tool.ruff.mccabe]
max-complexity = 10
Files excluded by .gitignore
are now ignored (#1234)
Ruff will now avoid checking files that are excluded by .ignore
, .gitignore
,
.git/info/exclude
, and global gitignore
files. This behavior is powered by the ignore
crate, and is applied in addition to Ruff's built-in exclude
system.
To disable this behavior, set respect-gitignore = false
in your pyproject.toml
file.
Note that hidden files (i.e., files and directories prefixed with a .
) are not ignored by
default.
Configuration files are now resolved hierarchically (#1190)
pyproject.toml
files are now resolved hierarchically, such that for each Python file, we find
the first pyproject.toml
file in its path, and use that to determine its lint settings.
See the documentation for more.