msg241506 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-04-19 10:21 |
asyncio/coroutines.py contains this code:
"""
_COROUTINE_TYPES = (types.GeneratorType, CoroWrapper)
def iscoroutine(obj):
"""Return True if obj is a coroutine object."""
return isinstance(obj, _COROUTINE_TYPES)
"""
In other places, it uses inspect.isgenerator() to do the same thing. This is inconsistent and should be changed. Code that wants to patch inspect.isgenerator() in order to support other generator types shouldn't also have to patch asyncio directly, and code that wants to support other generator types in asyncio shouldn't have to patch both places either.
|
msg241595 - (view) |
Author: Guido van Rossum (gvanrossum) *  |
Date: 2015-04-20 03:32 |
Uh, wait. Who's patching anything? That breaks the warranty.
|
msg241609 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-04-20 06:04 |
I was (silently) hoping that this patching would eventually not be necessary anymore because the one place (currently inspect.isgenerator()) would be adapted to check for the generator protocol rather than the generator type. But that was going to go into a separate ticket. :)
I'm not sure inspect.isgenerator() is the right place, though, as the module tends to deal with types, not protocols. I think the right fix overall would be a Generator ABC class that defines the protocol.
What Cython does now in order to make asyncio work with Cython compiled generators is this:
https://github.com/cython/cython/blob/4af42443bd37f6207143f8870904f780a65bb3e3/Cython/Utility/Generator.c#L824
https://github.com/cython/cython/blob/4af42443bd37f6207143f8870904f780a65bb3e3/Cython/Utility/Generator.c#L906
Aweful, but currently required due to the type check, which prevents objects that implement the generator protocol from being handled correctly by asyncio.
|
msg241676 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-04-20 19:37 |
I created issue 24018 for adding a Generator ABC to collections.abc.
|
msg243782 - (view) |
Author: Yury Selivanov (yselivanov) *  |
Date: 2015-05-21 19:36 |
This issue was resolved in https://github.com/python/asyncio/commit/3a09a93277afc2cdb43badf92a2c85c2789813f6.
|
msg244453 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-05-30 09:26 |
I found one more place in asyncio.coroutines, around line 190 in the coroutine() decorator:
if inspect.isgeneratorfunction(func):
coro = func
else:
@functools.wraps(func)
def coro(*args, **kw):
res = func(*args, **kw)
if isinstance(res, futures.Future) or inspect.isgenerator(res):
res = yield from res
return res
The wrapped isgenerator() check should be replaced by an Awaitable ABC check, e.g. this works for me:
"""
res = func(*args, **kw)
if isinstance(res, futures.Future) or inspect.isgenerator(res):
res = yield from res
+ else:
+ await_res = getattr(res, '__await__', None)
+ if await_res is not None:
+ res = yield from await_res()
return res
"""
|
msg244459 - (view) |
Author: Yury Selivanov (yselivanov) *  |
Date: 2015-05-30 14:16 |
What inspect.isgeneratorfunction(func) returns for Cython generators?
|
msg244474 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-05-30 17:56 |
Cython functions that return a generator aren't special in any way, so it's actually correct that isgeneratorfunction() always returns False. I mean, I could set the CO_COROUTINE flag on the code object that we fake, but that wouldn't help much as Cython's functions are not Python functions, so isfunction() will also return False for them and no-one would look at the flag in the first place. And finally, a Cython generator is not a Python generator (with byte code position etc.), so isgenerator() also correctly returns False.
At least iscoroutine() and iswaitable() will return True for Cython Coroutines now.
|
msg244479 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-05-30 18:18 |
Hmm, I just noticed that this only started failing when I disabled the asyncio module patching for 3.5beta1+, assuming that it now all works. A side effect of that was that also the inspect module is no longer patched into returning True from isgenerator() for Cython generators. While there might be code that fails when it tries to inspect Cython generators as Python generators (as the first don't have byte code), I think it's still the easiest fix to simply keep the inspect patching. That would then trigger the right code path here without changes in asyncio.
|
msg244482 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-05-30 18:28 |
... except that the returned object may not actually be a Cython generator but a GeneratorWrapper, now that the patch from issue 24316 is in place. :)
Then I guess we're back to the point where duck-typing and calling __await__() is a good idea.
|
msg244488 - (view) |
Author: Yury Selivanov (yselivanov) *  |
Date: 2015-05-30 20:01 |
Stefan,
> Then I guess we're back to the point where duck-typing and calling __await__() is a good idea.
Please test the attached patch. I agree this is a good idea.
|
msg244497 - (view) |
Author: Stefan Behnel (scoder) *  |
Date: 2015-05-30 21:00 |
Works for me.
Why is the AwaitableABC type check needed, in addition to looking up the relevant method? IIUC, the type check will just do a lookup of the same method if the type hasn't been registered as Awaitable explicitly.
|
msg244498 - (view) |
Author: Yury Selivanov (yselivanov) *  |
Date: 2015-05-30 21:01 |
> Why is the AwaitableABC type check needed, in addition to looking up the relevant method? IIUC, the type check will just do a lookup of the same method if the type hasn't been registered as Awaitable explicitly.
Because __await__ should be defined on the class, not on the method (as for all other magic methods in Python).
|
msg244507 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2015-05-31 01:02 |
New changeset b7b73029c825 by Yury Selivanov in branch '3.4':
Issue 24004: Support Awaitables (pep 492) in @asyncio.coroutine decorator
https://hg.python.org/cpython/rev/b7b73029c825
New changeset d1959cafc68c by Yury Selivanov in branch '3.5':
Issue 24004: Support Awaitables (pep 492) in @asyncio.coroutine decorator
https://hg.python.org/cpython/rev/d1959cafc68c
New changeset 6c53591d589b by Yury Selivanov in branch 'default':
Issue 24004: Support Awaitables (pep 492) in @asyncio.coroutine decorator
https://hg.python.org/cpython/rev/6c53591d589b
|
msg244508 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2015-05-31 01:05 |
New changeset 5f1e24f083c7 by Yury Selivanov in branch '3.5':
Issue 24004: Add a unittest for @asyncio.coroutine supporting Awaitables
https://hg.python.org/cpython/rev/5f1e24f083c7
New changeset 9d261141eb0c by Yury Selivanov in branch 'default':
Issue 24004: Add a unittest for @asyncio.coroutine supporting Awaitables
https://hg.python.org/cpython/rev/9d261141eb0c
|
msg244524 - (view) |
Author: Martin Panter (martin.panter) *  |
Date: 2015-05-31 08:00 |
Yury, your last change causes DeprecationWarning:
[ 38/398] test_asyncio
/media/disk/home/proj/python/cpython/Lib/test/test_asyncio/test_pep492.py:119: DeprecationWarning: Please use assertEqual instead.
self.assertEquals(coro.send(None), 'spam')
|
msg244543 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2015-05-31 15:29 |
New changeset 8296f0119f20 by Yury Selivanov in branch '3.5':
Issue 24004: Fix DeprecationWarning in a unittest
https://hg.python.org/cpython/rev/8296f0119f20
New changeset 60f5091cbfbf by Yury Selivanov in branch 'default':
Issue 24004: Fix DeprecationWarning in a unittest
https://hg.python.org/cpython/rev/60f5091cbfbf
|
msg244544 - (view) |
Author: Yury Selivanov (yselivanov) *  |
Date: 2015-05-31 15:29 |
Thanks, Martin!
|
|
Date |
User |
Action |
Args |
2022-04-11 14:58:15 | admin | set | github: 68192 |
2015-05-31 15:29:12 | yselivanov | set | messages:
+ msg244544 |
2015-05-31 15:29:05 | python-dev | set | messages:
+ msg244543 |
2015-05-31 08:00:02 | martin.panter | set | nosy:
+ martin.panter messages:
+ msg244524
|
2015-05-31 01:06:23 | yselivanov | set | status: open -> closed resolution: fixed |
2015-05-31 01:05:11 | python-dev | set | messages:
+ msg244508 |
2015-05-31 01:03:00 | python-dev | set | nosy:
+ python-dev messages:
+ msg244507
|
2015-05-30 21:01:56 | yselivanov | set | messages:
+ msg244498 |
2015-05-30 21:00:37 | scoder | set | status: pending -> open
messages:
+ msg244497 |
2015-05-30 20:02:03 | yselivanov | set | status: closed -> pending resolution: fixed -> (no value) |
2015-05-30 20:01:44 | yselivanov | set | files:
+ coroutine.patch keywords:
+ patch messages:
+ msg244488
|
2015-05-30 18:28:37 | scoder | set | messages:
+ msg244482 |
2015-05-30 18:18:52 | scoder | set | messages:
+ msg244479 |
2015-05-30 17:56:27 | scoder | set | messages:
+ msg244474 |
2015-05-30 14:16:40 | yselivanov | set | messages:
+ msg244459 |
2015-05-30 09:26:32 | scoder | set | messages:
+ msg244453 |
2015-05-21 19:36:03 | yselivanov | set | status: open -> closed versions:
- Python 3.4 messages:
+ msg243782
resolution: fixed stage: resolved |
2015-04-20 19:37:04 | scoder | set | messages:
+ msg241676 |
2015-04-20 06:04:55 | scoder | set | messages:
+ msg241609 |
2015-04-20 03:32:39 | gvanrossum | set | messages:
+ msg241595 |
2015-04-19 10:21:02 | scoder | create | |