Opened 13 months ago
Closed 13 months ago
#34983 closed Cleanup/optimization (fixed)
Deprecate django.utils.itercompat.is_iterable().
Reported by: | Nick Pope | Owned by: | Nick Pope |
---|---|---|---|
Component: | Utilities | Version: | dev |
Severity: | Normal | Keywords: | itercompat, is_iterable |
Cc: | Carsten Fuchs | Triage Stage: | Ready for checkin |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
While reviewing the PR for #10941, I saw that we are still using the django.utils.itercompat.is_iterable()
helper function.
It seems that this was originally added in ad077ccbc03f11c3db3951aab110ec98c5907729 to fix #5445 back in 2007. This was required because Jython 2.2 didn't have support for __iter__()
. Django's support for Jython was removed in 2017 in #28954.
As pointed out in this comment, is_iterable()
also handles the older fallback to determining if something is iterable by the presence of __getitem__()
as mentioned in Python's documentation:
Checking
isinstance(obj, Iterable)
detects classes that are registered asIterable
or that have an__iter__()
method, but it does not detect classes that iterate with the__getitem__()
method. The only reliable way to determine whether an object is iterable is to calliter(obj)
.
In practice, however, it is rare that iterables do not support __iter__()
and, for all uses of this internal function, it doesn't seem to be necessary.
I propose that we replace this function with the more standard isinstance(..., collections.abc.Iterable)
and deprecate it. (Although the function is private and undocumented, it is likely used in the wild and other past functions from this module have been through a deprecation process.)
Change History (12)
comment:1 by , 13 months ago
Has patch: | set |
---|
comment:2 by , 13 months ago
Thanks for the ticket. Have you checked how it looks like for PyPy? I'm afraid removing it will make things worse for PyPy. For example ChoiceIterator
s don't use is_iterable()
and they seem to be broken on PyPy:
./runtests.py model_fields .... django.core.management.base.SystemCheckError: SystemCheckError: System check identified some issues: ERRORS: model_fields.Choiceful.choices_from_callable: (fields.E004) 'choices' must be a mapping (e.g. a dictionary) or an iterable (e.g. a list or tuple). model_fields.Choiceful.with_choices_dict: (fields.E005) 'choices' must be a mapping of actual values to human readable names or an iterable containing (actual value, human readable name) tuples. model_fields.Choiceful.with_choices_nested_dict: (fields.E005) 'choices' must be a mapping of actual values to human readable names or an iterable containing (actual value, human readable name) tuples. model_fields.Whiz.c: (fields.E005) 'choices' must be a mapping of actual values to human readable names or an iterable containing (actual value, human readable name) tuples. model_fields.WhizDelayed.c: (fields.E005) 'choices' must be a mapping of actual values to human readable names or an iterable containing (actual value, human readable name) tuples. model_fields.WhizIter.c: (fields.E005) 'choices' must be a mapping of actual values to human readable names or an iterable containing (actual value, human readable name) tuples. System check identified 6 issues (0 silenced).
comment:3 by , 13 months ago
I'll dig into the issues around the choices stuff some more, but in the general simple case it seems like it should be fine:
❯ pypy3 Python 3.10.13 (f1607341da97ff5a1e93430b6e8c4af0ad1aa019, Oct 02 2023, 19:25:21) [PyPy 7.3.13 with GCC 13.2.1 20230801] on linux Type "help", "copyright", "credits" or "license" for more information. >>>> from collections.abc import Iterable >>>> class A: .... def __iter__(self): .... pass .... >>>> isinstance(A(), Iterable) True >>>>
follow-up: 6 comment:4 by , 13 months ago
It seems that 500e01073adda32d5149624ee9a5cb7aa3d3583f introduces a big regression on PyPy, e.g.
- before we had 3 failures in
forms_tests
, afterfailures=13, errors=32
, - before we had no failures in
model_fields
, after system check identifies issues, etc.
comment:5 by , 13 months ago
Cc: | added |
---|
comment:6 by , 13 months ago
Replying to Mariusz Felisiak:
It seems that 500e01073adda32d5149624ee9a5cb7aa3d3583f introduces a big regression on PyPy, e.g.
While doing the review for that PR, I did check PyPy compatibility. Is that something we should be doing? Thanks!
comment:7 by , 13 months ago
So 3.10 support in PyPy is still quite new - less than 6 months old.
It looks like we've found a bug for which I've opened an issue: https://foss.heptapod.net/pypy/pypy/-/issues/4036
That is not related to this ticket, however - what we're looking at here seems to work fine.
comment:9 by , 13 months ago
Patch needs improvement: | set |
---|
comment:10 by , 13 months ago
Patch needs improvement: | unset |
---|
comment:11 by , 13 months ago
Triage Stage: | Accepted → Ready for checkin |
---|
PR