[issue23934] inspect.signature reporting () for all builtin extension types
Changes by Yury Selivanov yseliva...@gmail.com: -- resolution: - fixed stage: test needed - resolved status: open - closed ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
Roundup Robot added the comment: New changeset e59966bb6de5 by Yury Selivanov in branch '3.5': Issue #23934: Fix inspect.signature to fail correctly for builtin types. https://hg.python.org/cpython/rev/e59966bb6de5 New changeset 19e0ffdd1daa by Yury Selivanov in branch 'default': Issue #23934: Fix inspect.signature to fail correctly for builtin types. https://hg.python.org/cpython/rev/19e0ffdd1daa -- nosy: +python-dev ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
Nick Coghlan added the comment: The main missing piece now is test fodder for this error. The simplest near-term option would be to use one of the actual builtins, such as 'str'. In Python 3.5+ we could borrow the matmulType that was added to _testcapi to test the @ operator, future-proofing the test against eventual conversion of the builtin types to support Argument Clinic. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
James Powell added the comment: See attached patch for unittest. For 3.4, test that inspect.signature(str) raises TypeError. For 3.5, this can be improved to use _testcapi.matmulType -- Added file: http://bugs.python.org/file38949/issue_23934-test.patch ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
Yury Selivanov added the comment: The patch looks good. -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
James Powell added the comment: Discussed with Nick Coghlan. See attached patch to return `signature(object)` only if both `__new__` and `__init__` are shared with `object`. Otherwise, raise TypeError indicating built-in types not supported. https://docs.python.org/3/library/inspect.html#inspect.signature Raises ValueError if no signature can be provided, and TypeError if that type of object is not supported. -- keywords: +patch Added file: http://bugs.python.org/file38936/issue_23934.patch ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
New submission from Nick Coghlan:
inspect.signature isn't currently handling builtin extension types correctly
- these show up as having neither __new__ *nor* __init__ as pure Python
callables, so the inspect.signature logic falls through into a currently
unhandled case.
_testcapi isn't currently exporting an extension type as docstring
introspection fodder, only callables, so test_inspect didn't pick up the
problem. The problem can be seen with builtin types like str:
import inspect
inspect.signature(str)
inspect.Signature object at 0x7fb81d44e518
print(inspect.signature(str))
()
Expected behaviour would be to throw a ValueError as with builtin callables
without a signature:
import _testcapi
import inspect
inspect.signature(_testcapi.docstring_no_signature)
Traceback (most recent call last):
File stdin, line 1, in module
File /home/ncoghlan/devel/py3k/Lib/inspect.py, line 2830, in signature
return Signature.from_callable(obj)
File /home/ncoghlan/devel/py3k/Lib/inspect.py, line 2586, in from_callable
return _signature_from_callable(obj, sigcls=cls)
File /home/ncoghlan/devel/py3k/Lib/inspect.py, line 2064, in
_signature_from_callable
skip_bound_arg=skip_bound_arg)
File /home/ncoghlan/devel/py3k/Lib/inspect.py, line 1984, in
_signature_from_builtin
raise ValueError(no signature found for builtin {!r}.format(func))
ValueError: no signature found for builtin built-in function
docstring_no_signature
--
messages: 240649
nosy: james, larry, ncoghlan, yselivanov
priority: normal
severity: normal
stage: test needed
status: open
title: inspect.signature reporting () for all builtin extension types
type: behavior
versions: Python 3.4, Python 3.5
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue23934
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
Larry Hastings added the comment: Would this qualify as an easy bug we could throw to a sprinter? -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23934] inspect.signature reporting () for all builtin extension types
Nick Coghlan added the comment: James Powell is currently looking into this at the PyCon sprints. The key problem appears to be the check that assumes obj.__init__ is object.__init__ indicates that there's no user defined init or new method, when in fact a builtin or extension type that only overrides __new__ would also pass that check (since the earlier check for __new__ returns None if the method implementation isn't a pure Python callable) -- ___ Python tracker rep...@bugs.python.org http://bugs.python.org/issue23934 ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
