-
-
Notifications
You must be signed in to change notification settings - Fork 30.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
bpo-29679: Implement @contextlib.asynccontextmanager #360
Conversation
@JelleZijlstra, thanks for your PR! By analyzing the history of the files in this pull request, we identified @gvanrossum, @brettcannon and @Yhg1s to be potential reviewers. |
Also adding @ncoghlan who I believe maintains contextlib (although I can't find module maintainers listed anywhere in the devguide or the repo). Not sure what prompted the error in mention-bot's first message. |
@JelleZijlstra The maintainer list is at https://docs.python.org/devguide/experts.html#experts (you can also just start typing the module name into the nosy list field on bugs.python.org and it will autopopulate the dropdown with the relevant usernames) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
General concept and approach looks good to me, just some specific feedback on particular aspects of the implementation and test layout.
Lib/contextlib.py
Outdated
@@ -54,8 +54,9 @@ def inner(*args, **kwds): | |||
return inner | |||
|
|||
|
|||
class _GeneratorContextManager(ContextDecorator, AbstractContextManager): | |||
"""Helper for @contextmanager decorator.""" | |||
class _GeneratorContextManagerBase(ContextDecorator): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ContextDecorator is designed to work with __call__
rather than __await__
, so having it also applied to asynccontextmanager
doesn't look right to me.
So I'd suggest moving the ContextDecorator inheritance and the _recreate_cm
method down into _GeneratorContextManager
and only keeping the __init__
method common between the two.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You're right, my implementation didn't actually work as a decorator. (Although it's because ContextDecorator uses with
instead of async with
, not because of anything to do with __call__
.)
We could add a separate AsyncContextDecorator
implementation, but I don't currently have a good use for that.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I actually consider making _GeneratorContextManager inherit from ContextDecorator a design error on my part (that's why _recreate_cm is still a private API: https://bugs.python.org/issue11647#msg161113 )
So simply not supporting an equivalent for async context managers is entirely fine by me :)
Lib/contextlib.py
Outdated
@@ -160,6 +203,40 @@ def helper(*args, **kwds): | |||
return helper | |||
|
|||
|
|||
def asynccontextmanager(func): | |||
"""@contextmanager decorator. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Missing async
in the docstring.
Lib/test/test_contextlib.py
Outdated
@@ -1,5 +1,6 @@ | |||
"""Unit tests for contextlib.py, and other context managers.""" | |||
|
|||
import asyncio |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd prefer for other implementations to be able to readily test the rest of contextlib without requiring a working asyncio implementation, so it would be good to split these new tests out to a separate Lib/test_contextlib_async.py
file.
Thanks for reviewing so quickly! I've pushed fixes for the issues you raised. |
The PR itself looks good to me now, but I'm going to place it on hold for the moment pending further consideration of whether we should add this to |
@1st1 is someone who might be interested in this, see for example the discussion of |
OK, based on the python-dev discussion, I'm happy with the idea of proceeding with this basic API design: http://bugs.python.org/issue29679#msg288781 However, those code coverage results are suspicious, as they suggest that either the async-based tests aren't actually running properly, or else the coverage report isn't capturing their execution correctly. Could you take another look at those tests and inject some deliberate failures (I like |
That's strange. When I change the code locally to make one of the tests fail, they definitely do fail. (I ran The detailed Travis results show that |
Found the issue and pushed a fix. The problem was that test_asyncio closes the default event loop, so if test_asyncio and test_contextlib_async are run consecutively in the same process, the contextlib_async tests that rely on get_event_loop() returning a usable loop fail. I fixed the problem by instead creating a new event loop. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Noted several lines from the diff coverage report that the new test cases should really be hitting. The other uncovered lines all step from contextlib
being imported being the coverage data starts being collected, so there isn't a lot to be done about that.
Feel free to ask for more info if it isn't clear how to craft a test case that covers the commented on lines.
try: | ||
return await self.gen.__anext__() | ||
except StopAsyncIteration: | ||
raise RuntimeError("generator didn't yield") from None |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Diff coverage shows a missing test case for this line.
except StopAsyncIteration: | ||
return | ||
else: | ||
raise RuntimeError("generator didn't stop") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Missing test case here as well.
Lib/contextlib.py
Outdated
raise RuntimeError("generator didn't stop") | ||
else: | ||
if value is None: | ||
value = type() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You won't be able to hit this line via the async with
syntax, but it can be exercised by awaiting __aexit__
directly with a non-normalised exception (i.e. only the exception type, with both the exception value and the traceback as None)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We could also write a small C helper to do that, but your idea is much better.
Lib/contextlib.py
Outdated
except RuntimeError as exc: | ||
if exc is value: | ||
return False | ||
if exc.__cause__ is value: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Huh, this (and its counterpart in _GeneratorContextManager
) is buggy, since it isn't checking that value
was the appropriate kind of exception to start with: http://bugs.python.org/issue29692
Don't fix the synchronous version in this PR, but do add a test case for the async version that hits this line (throw StopAsyncIteration
from inside an async with statement and then catch it again outside), and also one that chains some other exception type with RuntimeError and make sure that still gets chained correctly.
Lib/contextlib.py
Outdated
raise | ||
except: | ||
if sys.exc_info()[1] is not value: | ||
raise |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hitting this line means adding a test case that replaces the thrown in exception with an entirely unrelated one that is neither StopAsyncIteration
nor RuntimeError
Lib/test/test_contextlib_async.py
Outdated
def _async_test(func): | ||
"""Decorator to turn an async function into a test case.""" | ||
def wrapper(*args, **kwargs): | ||
loop = asyncio.new_event_loop() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You need to close the loop (1), and make sure it's the default loop (2).
Please rewrite to:
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
try:
loop.run_until_complete(coro)
finally:
loop.close()
asyncio.set_event_loop(None)
Lib/test/test_contextlib_async.py
Outdated
|
||
def _async_test(func): | ||
"""Decorator to turn an async function into a test case.""" | ||
def wrapper(*args, **kwargs): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also, I know it's just a test, but please add @functools.wraps(func)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add a wraps
decorator.
Doc/library/contextlib.rst
Outdated
.. decorator:: asynccontextmanager | ||
|
||
Similar to :func:`~contextlib.contextmanager`, but works with | ||
:term:`coroutines <coroutine>`. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"works with coroutine" isn't really descriptive. I'd rephrase to something akin to "but creates asynchronous context managers".
Lib/contextlib.py
Outdated
class _GeneratorContextManager(ContextDecorator, AbstractContextManager): | ||
"""Helper for @contextmanager decorator.""" | ||
class _GeneratorContextManagerBase: | ||
"""Shared functionality for the @contextmanager and @asynccontextmanager |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
First line of docstring must be a single sentence. Just make it
"""Shared functionality for the @contextmanager and @asynccontextmanager."""
Thanks! Fixed the comment. |
If you'd like, you can add a note to Docs/whatsnew/3.7.rst about the addition, otherwise I'll add it when I update Misc/NEWS prior to merging :) |
Doc/library/contextlib.rst
Outdated
This function is a :term:`decorator` that can be used to define a factory | ||
function for :keyword:`async with` statement asynchronous context managers, | ||
without needing to create a class or separate :meth:`__aenter__` and | ||
:meth:`__aexit__` methods. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd add that the decorator expects to be applied to asynchronous generator functions.
Doc/whatsnew/3.7.rst
Outdated
unittest.mock | ||
------------- | ||
|
||
The :const:`~unittest.mock.sentinel` attributes now preserve their identity | ||
when they are :mod:`copied <copy>` or :mod:`pickled <pickle>`. | ||
(Contributed by Serhiy Storchaka in :issue:`20804`.) | ||
|
||
xmlrpc.server |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd leave reordering out of this PR. It should be done in a separate commit.
I've left another couple of comments in the review. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM. FWIW I'm really glad the tests didn't expose any bugs in asynchronous generators :)
Created the same package that implements asynccontextmanager decorator on these weekends, and wanted to push it into python. And here it is! Any thoughts in which python release this will be included? |
|
Correct. @ncoghlan do you want me to merge this PR? |
@1st1 Go ahead, thanks :) For folks interested in 3.5 and 3.6 support, I opened an issue against contextlib2 to discuss the options for handling that: jazzband/contextlib2#12 |
Not sure what happened to the Appveyor build, the failures don't look related to contextlib. |
Is there anything blocking this from being merged? There were a few merge conflicts already. |
I'll merge it when it passes the checks. Thanks! |
Thanks! |
@JelleZijlstra |
Yes, sounds good. I can do that. |
Implements: - python/typing#438 - python/cpython#360 python/cpython#1412, which adds contextlib.AbstractAsyncContextManager, has not yet been merged.
) Implements: - python/typing#438 - python/cpython#360 Note that python/cpython#1412, which adds contextlib.AbstractAsyncContextManager, has not yet been merged.
#360) * Remove 'Automerge-Triggered-By' text when removing the automerge label * Add more tests * Add test case for the scenario that the trailer text does not exist
No description provided.