-
-
Notifications
You must be signed in to change notification settings - Fork 30.7k
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-38337: Change getattr to inspect.getattr_static #16521
Conversation
Hello, and thanks for your contribution! I'm a bot set up to make sure that the project can legally accept this contribution by verifying everyone involved has signed the PSF contributor agreement (CLA). CLA MissingOur records indicate the following people have not signed the CLA: @jnsdrtlf For legal reasons we need all the people listed to sign the CLA before we can look at your contribution. Please follow the steps outlined in the CPython devguide to rectify this issue. If you have recently signed the CLA, please wait at least one business day We also demand... A SHRUBBERY! You can check yourself to see if the CLA has been received. Thanks again for the contribution, we look forward to reviewing it! |
should we add a test case ? |
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.
Thanks @jnsdrtlf, and welcome to CPython! 😎
I have a minor change to suggest for the NEWS entry, but otherwise this looks like a good improvement.
I also agree that adding regression test is probably a good idea.
Misc/NEWS.d/next/Library/2019-10-01-12-23-37.bpo-38337.QmLSXW.rst
Outdated
Show resolved
Hide resolved
Actually, from looking at the test failures... I'm not sure if this is the desired behavior (and the docs are ambiguous). Maybe @1st1 can shed some light on the "correct" behavior of this function? |
Since |
Okay then. So this probably doesn't need a new test, but the ~4 failing ones should be updated! |
I think it needs a new test. Basically an object with a property and a test ensuring that the property was discovered by getmembers, but the getter code wasn't run. |
As suggested by @brandtbucher Co-Authored-By: Brandt Bucher <brandtbucher@gmail.com>
Sounds good! I will get back to work and write a test later today. |
Make sure inspect.getmembers won't call properties or other dynamic attributes
The test @types.DynamicClassAttribute
def eggs(self):
return 'spam' to be called. It asserts whether There is also some wierd behaviour when overwriting Failing tests:
|
Lib/test/test_inspect.py
Outdated
@@ -1191,6 +1191,27 @@ def f(self): | |||
self.assertIn(('f', b.f), inspect.getmembers(b)) | |||
self.assertIn(('f', b.f), inspect.getmembers(b, inspect.ismethod)) | |||
|
|||
def test_getmembers_static(self): | |||
class Foo: | |||
def __init__(bar): |
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.
Use def __init__(self, bar):
please.
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.
Oups 🤦♂ Fixed in commit b415224
This comment has been minimized.
This comment has been minimized.
Another inconsistent behaviour:
However, the following code does not return any properties: class Foo:
def __init__(self, data):
self._bar = data
@property
def bar(self):
print('--BAR--')
return self._bar
inspect.getmembers(Foo('test'), inspect.isdatadescriptor) Result
Expected (new result) [('__class__', <attribute '__class__' of 'object' objects>), ('__dict__', <attribute '__dict__' of 'Foo' objects>), ('__weakref__', <attribute '__weakref__' of 'Foo' objects>), ('bar', <property object at 0x...>)] Is this correct so far? I am still confused about how |
Additionally, class Foo:
pass
inspect.getattr_static(Foo('test'), '__class__') returns class Foo:
pass
# Will return [] if getattr_static is used
inspect.getmembers(Foo('test'), inspect.isclass) |
@@ -347,7 +347,10 @@ def getmembers(object, predicate=None): | |||
# like calling their __get__ (see bug #1785), so fall back to | |||
# looking in the __dict__. | |||
try: | |||
value = getattr_static(object, key) | |||
if isinstance(getattr_static(object, key), property): |
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.
Why can't we always use getattr_static
? If there's a reason for not using it always, please describe that in a comment.
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.
This is related to my latest comment. Sorry, I didn't make this clear:
getattr_static(...,'__class__')
returns <attribute '__class__' of 'object' objects>
whereas getattr
would return <class '__main__.Foo'>
. What should getmembers return? Using getattr_static
for every attribute doesn't seem to be the solution. It works fine with methods, functions and properties, but some edge cases seem to require getattr
(such as __class__
). It is not clear what the expected behaviour is.
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.
Got it, well, describe this in a comment :)
In 3ac2093, I've changed
This feels more like a temporary fix - could be reverted to the first fix I've committed - although I am just not sure which is better. At this point, I am not sure what the actual expected behaviour should be. Someone should review this and the test case that is currently failing, as it expects a different behaviour (return the actual value of the property). |
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.
Jonas, I thought about this PR for some time and I think we should reject it. Modifying getmembers
in any way would break backwards compatibility in a very non-obvious way for many users. The ship to fix it has sailed.
We can add a new function getmembers_static
that would use getattr_static
instead of getattr
. It must be a new distinct function, though.
A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated. Once you have made the requested changes, please leave a comment on this pull request containing the phrase |
@1st1 I share your thoughts on that. It does not seem to be obvious, what As for introducing a function
Is there a better place to discuss such questions? |
I think it would be better if you could post this to the python-ideas mailing list. I've very limited time and would hate to get this thing staled because of me. |
Or post it here: https://discuss.python.org/c/ideas |
I added a link to the discussion on the bug tracker. |
This would be very helpful, running into the same issue with inspect.getmembers evaluating @Property decorated methods. getmembers_static would solve it. |
When calling
inspect.getmembers
on a class that has a property (@property
), the property will be called by thegetattr
call ingetmembers
.Example:
Result:
Expected (new result):
https://bugs.python.org/issue38337