A highly opinionated flake8 plugin for problems related to Trio, AnyIO, or asyncio.
This can include anything from outright bugs, to pointless/dead code, to likely performance issues, to minor points of idiom that might signal a misunderstanding.
It may well be too noisy for anyone with different opinions, that's OK.
Pairs well with flake8-bugbear.
Some checks are incorporated into ruff.
This plugin was previously known as flake8-trio, and there was a separate small plugin known as flake8-async for asyncio. But this plugin was a superset of the checks in flake8-async, and support for anyio was added, so it's now named flake8-async to more properly convey its usage. At the same time all error codes were renamed from TRIOxxx to ASYNCxxx, as was previously used by the old flake8-async.
pip install flake8-async
-
ASYNC100: A
with [trio/anyio].fail_after(...):
orwith [trio/anyio].move_on_after(...):
context does not contain anyawait
statements. This makes it pointless, as the timeout can only be triggered by a checkpoint. This check also allowsyield
statements, since checkpoints can happen in the caller we yield to. -
ASYNC101:
yield
inside a trio/anyio nursery or cancel scope is only safe when implementing a context manager - otherwise, it breaks exception handling. -
ASYNC102: It's unsafe to await inside
finally:
orexcept BaseException/trio.Cancelled/anyio.get_cancelled_exc_class()/asyncio.exceptions.CancelledError
unless you use a shielded cancel scope with a timeout. This is currently not able to detect asyncio shields. -
ASYNC103:
except BaseException/trio.Cancelled/anyio.get_cancelled_exc_class()/asyncio.exceptions.CancelledError
, or a bareexcept:
with a code path that doesn't re-raise. If you don't want to re-raiseBaseException
, add a separate handler fortrio.Cancelled
/anyio.get_cancelled_exc_class()
/asyncio.exceptions.CancelledError
before. -
ASYNC104:
trio.Cancelled
/anyio.get_cancelled_exc_class()
/asyncio.exceptions.CancelledError
/BaseException
must be re-raised. The same as ASYNC103, except specifically triggered onreturn
or a different exception being raised. -
ASYNC105: Calling a trio async function without immediately
await
ing it. This is only supported with trio functions, but you can get similar functionality with a type-checker. -
ASYNC106:
trio
/anyio
/asyncio
must be imported withimport trio
/import anyio
/import asyncio
for the linter to work. -
ASYNC109: Async function definition with a
timeout
parameter - use[trio/anyio].[fail/move_on]_[after/at]
instead. -
ASYNC110:
while <condition>: await [trio/anyio].sleep()
should be replaced by a[trio/anyio].Event
. -
ASYNC111: Variable, from context manager opened inside nursery, passed to
start[_soon]
might be invalidly accessed while in use, due to context manager closing before the nursery. This is usually a bug, and nurseries should generally be the inner-most context manager. -
ASYNC112: Nursery body with only a call to
nursery.start[_soon]
and not passing itself as a parameter can be replaced with a regular function call. -
ASYNC113: Using
nursery.start_soon
in__aenter__
doesn't wait for the task to begin. Consider replacing withnursery.start
. -
ASYNC114: Startable function (i.e. has a
task_status
keyword parameter) not in--startable-in-context-manager
parameter list, please add it so ASYNC113 can catch errors when using it. -
ASYNC115: Replace
[trio/anyio].sleep(0)
with the more suggestive[trio/anyio].lowlevel.checkpoint()
. -
ASYNC116:
[trio/anyio].sleep()
with >24 hour interval should usually be[trio/anyio].sleep_forever()
. -
ASYNC118: Don't assign the value of
anyio.get_cancelled_exc_class()
to a variable, since that breaks linter checks and multi-backend programs. -
ASYNC119:
yield
in context manager in async generator is unsafe, the cleanup may be delayed untilawait
is no longer allowed. We strongly encourage you to read PEP-533 and useasync with aclosing(...)
, or better yet avoid async generators entirely (see ASYNC900) in favor of context managers which return an iterable channel/queue.
Note: 22X, 23X and 24X has not had asyncio-specific suggestions written.
-
ASYNC200: User-configured error for blocking sync calls in async functions. Does nothing by default, see
async200-blocking-calls
for how to configure it. -
ASYNC210: Sync HTTP call in async function, use
httpx.AsyncClient
. This and the other ASYNC21x checks look for usage ofurllib3
andhttpx.Client
, and recommend usinghttpx.AsyncClient
as that's the largest http client supporting anyio/trio. -
ASYNC211: Likely sync HTTP call in async function, use
httpx.AsyncClient
. Looks forurllib3
method calls on pool objects, but only matching on the method signature and not the object. - ASYNC212: Blocking sync HTTP call on httpx object, use httpx.AsyncClient.
-
ASYNC220: Sync process call in async function, use
await nursery.start([trio/anyio].run_process, ...)
.asyncio
users can useasyncio.create_subprocess_[exec/shell]
. -
ASYNC221: Sync process call in async function, use
await [trio/anyio].run_process(...)
.asyncio
users can useasyncio.create_subprocess_[exec/shell]
. -
ASYNC222: Sync
os.*
call in async function, wrap inawait [trio/anyio].to_thread.run_sync()
.asyncio
users can useasyncio.loop.run_in_executor
. -
ASYNC230: Sync IO call in async function, use
[trio/anyio].open_file(...)
.asyncio
users need to use a library such as aiofiles, or switch to anyio. -
ASYNC231: Sync IO call in async function, use
[trio/anyio].wrap_file(...)
.asyncio
users need to use a library such as aiofiles, or switch to anyio. -
ASYNC232: Blocking sync call on file object, wrap the file object in
[trio/anyio].wrap_file()
to get an async file object. -
ASYNC240: Avoid using
os.path
in async functions, prefer using[trio/anyio].Path
objects.asyncio
users should consider aiopath or anyio. -
ASYNC250: Builtin
input()
should not be called from async function. Wrap in[trio/anyio].to_thread.run_sync()
orasyncio.loop.run_in_executor()
. -
ASYNC251:
time.sleep(...)
should not be called from async function. Use[trio/anyio/asyncio].sleep(...)
.
-
ASYNC900: Async generator without
@asynccontextmanager
not allowed. You might want to enable this on a codebase since async generators are inherently unsafe and cleanup logic might not be performed. See #211 and https://discuss.python.org/t/using-exceptiongroup-at-anthropic-experience-report/20888/6 for discussion. -
ASYNC910: Exit or
return
from async function with no guaranteed checkpoint or exception since function definition. You might want to enable this on a codebase to make it easier to reason about checkpoints, and make the logic of ASYNC911 correct. -
ASYNC911: Exit,
yield
orreturn
from async iterable with no guaranteed checkpoint since possible function entry (yield or function definition) Checkpoints areawait
,async for
, andasync with
(on one of enter/exit).
- TRIOxxx: All error codes are now renamed ASYNCxxx
- TRIO107: Renamed to TRIO910
- TRIO108: Renamed to TRIO911
-
TRIO117: Don't raise or catch
trio.[NonBase]MultiError
, prefer[exceptiongroup.]BaseExceptionGroup
.MultiError
was removed in trio==0.24.0.
pip install flake8 flake8-async
flake8 .
If you use pre-commit, you can use it with flake8-async by
adding the following to your .pre-commit-config.yaml
:
minimum_pre_commit_version: '2.9.0'
repos:
- repo: https://github.com/python-trio/flake8-async
rev: 23.2.5
hooks:
- id: flake8-async
# args: [--enable=ASYNC, --disable=ASYNC9, --autofix=ASYNC]
This is often considerably faster for large projects, because pre-commit
can avoid running flake8-async
on unchanged files.
Afterwards, run
pip install pre-commit flake8-async
pre-commit run .
If inside a git repository, running without arguments will run it against all *.py
files in the repository.
pip install flake8-async
flake8-async
flake8-async --autofix=ASYNC
flake8-async my_python_file.py
flake8-async **/*.py
You can configure flake8
with command-line options,
but we prefer using a config file. The file needs to start with a section marker [flake8]
and the following options are then parsed using flake8's config parser, and can be used just like any other flake8 options.
Note that it's not currently possible to use a configuration file when running flake8-async
standalone.
Error codes with more than three letters are not possible to ignore
in config files since flake8>=6, as flake8 tries to validate correct configuration with a regex. We have decided not to conform to this, as it would be a breaking change for end-users requiring them to update noqa
s and configurations, we think the ASYNC
code is much more readable than e.g. ASYxxx
, and ruff does not enforce such a limit. The easiest option for users hitting this error is to instead use the --disable
option as documented below. See further discussion and other workarounds in #230
Comma-separated list of error codes to enable, similar to flake8 --select but is additionally more performant as it will disable non-enabled visitors from running instead of just silencing their errors.
Comma-separated list of error codes to disable, similar to flake8 --ignore but is additionally more performant as it will disable non-enabled visitors from running instead of just silencing their errors.
Comma-separated list of error-codes to enable autofixing for if implemented. Requires running as a standalone program. Pass --autofix=ASYNC
to enable all autofixes.
Whether to also print an error message for autofixed errors.
Change the default library to be anyio instead of trio. If trio is imported it will assume both are available and print suggestions with [anyio/trio].
Comma-separated list of decorators to disable checkpointing checks for, turning off ASYNC910 and ASYNC911 warnings for functions decorated with any decorator matching any in the list. Matching is done with fnmatch. Defaults to disabling for asynccontextmanager
.
Decorators-to-match must be identifiers or dotted names only (not PEP-614 expressions), and will match against the name only - e.g. foo.bar
matches foo.bar
, foo.bar()
, and foo.bar(args, here)
, etc.
For example:
no-checkpoint-warning-decorators =
mydecorator,
mydecoratorpackage.checkpointing_decorators.*,
ign*,
*.ignore,
Comma-separated list of methods which should be used with .start()
when opening a context manager,
in addition to the default trio.run_process
, trio.serve_tcp
, trio.serve_ssl_over_tcp
, and
trio.serve_listeners
. Names must be valid identifiers as per str.isidentifier()
. For example:
startable-in-context-manager =
myfun,
myfun2,
Comma-separated list of pairs of values separated by ->
(optional whitespace stripped), where the first is a pattern for a call that should raise an error if found inside an async function, and the second is what should be suggested to use instead. It uses fnmatch as per no-checkpoint-warning-decorators
for matching. The part after ->
is not used by the checker other than when printing the error, so you could add extra info there if you want.
The format of the error message is User-configured blocking sync call {0} in async function, consider replacing with {1}.
, where {0}
is the pattern the call matches and {1}
is the suggested replacement.
Example:
async200-blocking-calls =
my_blocking_call -> async.alternative,
module.block_call -> other_function_to_use,
common_error_call -> alternative(). But sometimes you should use other_function(). Ask joe if you're unsure which one,
dangerous_module.* -> corresponding function in safe_module,
*.dangerous_call -> .safe_call()
Specified patterns must not have parentheses, and will only match when the pattern is the name of a call, so given the above configuration
async def my_function():
my_blocking_call() # this would raise an error
x = my_blocking_call(a, b, c) # as would this
y = my_blocking_call # but not this
y() # or this
[my_blocking_call][0]() # nor this
def my_blocking_call(): # it's also safe to use the name in other contexts
...
arbitrary_other_function(my_blocking_call=None)