gh-113978: Ignore warnings on text completion inside REPL

[WolframAlph]

gh-113978

[sobolevn]

Maybe instead we should do this?

» git patch
diff --git Lib/rlcompleter.py Lib/rlcompleter.py
index 206d6fb511c..62fd2c7fdec 100644
--- Lib/rlcompleter.py
+++ Lib/rlcompleter.py
@@ -176,8 +176,9 @@ def attr_matches(self, text):
                 if (word[:n] == attr and
                     not (noprefix and word[:n+1] == noprefix)):
                     match = "%s.%s" % (expr, word)
+                    import types
                     if isinstance(getattr(type(thisobject), word, None),
-                                  property):
+                                  (property, types.GetSetDescriptorType)):
                         # bpo-44752: thisobject.word is a method decorated by
                         # `@property`. What follows applies a postfix if
                         # thisobject.word is callable, but know we know that
                        

It also solves the issue:

[WolframAlph]

Not really sure. My intention was to eliminate all possible warnings during completion (now and in future) no matter where they come from. Maybe it's the only possible path where it can happen but I am not sure since I never dealt with such things. Your solution looks cleaner to me and I believe you have broader context so I leave it up to you to decide.

[sobolevn]

My thoughs: properties should not evaluate when auto-complete works, so I think that adding an ignore just hides this problem.

So, we should really address the root of this problem: do not execute .co_lnotab property on code objects (and similar types as well).

[WolframAlph]

properties should not evaluate when auto-complete works

I agree. But is this true only for built-in objects and their properties or user defined as well?
Consider following piece of code:

class D:
    def __get__(self, _, __):
        print("evaluated")

class C:
    prop = D()

trying to get suggestions for C. evaluates prop and prints the message.

[sobolevn]

We can also try to replace getattr with inspect.getattr_static

[WolframAlph]

Some things depend on being evaluated. Using inspect.getattr_static instead of getattr breaks tests.

cpython/Lib/test/test_rlcompleter.py

Lines 51 to 85 in 77b45fa

	def test_attr_matches(self):
	# test with builtins namespace
	self.assertEqual(self.stdcompleter.attr_matches('str.s'),
	['str.{}('.format(x) for x in dir(str)
	if x.startswith('s')])
	self.assertEqual(self.stdcompleter.attr_matches('tuple.foospamegg'), [])
	expected = sorted({'None.%s%s' % (x,
	'()' if x == '__init_subclass__'
	else '' if x == '__doc__'
	else '(')
	for x in dir(None)})
	self.assertEqual(self.stdcompleter.attr_matches('None.'), expected)
	self.assertEqual(self.stdcompleter.attr_matches('None._'), expected)
	self.assertEqual(self.stdcompleter.attr_matches('None.__'), expected)
	# test with a customized namespace
	self.assertEqual(self.completer.attr_matches('CompleteMe.sp'),
	['CompleteMe.spam'])
	self.assertEqual(self.completer.attr_matches('Completeme.egg'), [])
	self.assertEqual(self.completer.attr_matches('CompleteMe.'),
	['CompleteMe.mro()', 'CompleteMe.spam'])
	self.assertEqual(self.completer.attr_matches('CompleteMe._'),
	['CompleteMe._ham'])
	matches = self.completer.attr_matches('CompleteMe.__')
	for x in matches:
	self.assertTrue(x.startswith('CompleteMe.__'), x)
	self.assertIn('CompleteMe.__name__', matches)
	self.assertIn('CompleteMe.__new__(', matches)
	with patch.object(CompleteMe, "me", CompleteMe, create=True):
	self.assertEqual(self.completer.attr_matches('CompleteMe.me.me.sp'),
	['CompleteMe.me.me.spam'])
	self.assertEqual(self.completer.attr_matches('egg.s'),
	['egg.{}('.format(x) for x in dir(str)
	if x.startswith('s')])

Things like __init_subclass__, __class__, mro, etc...

[miss-islington-app]

Thanks @WolframAlph for the PR, and @ambv for merging it 🌮🎉.. I'm working now to backport this PR to: 3.13.
🐍🍒⛏🤖

[bedevere-app]

GH-119429 is a backport of this pull request to the 3.13 branch.