Fixed false positive of no-member for methods of namedtuples instances

[yushao2]

Steps

• Add a ChangeLog entry describing what your PR does.
• If it's a new feature or an important bug fix, add a What's New entry in
  doc/whatsnew/<current release.rst>. (not sure if important bugfix)
• Write a good description on what the PR does.

Description

From investigation, it seems like _replace() yields a NamedTuple instance and .geturl() is not a valid attribute -- therefore the warning is triggered.

>>> pprint(instance.as_string())
('\n'
 '\n'
 'class ParseResult(tuple):\n'
 '    __slots__ = ()\n'
 "    _fields = ['scheme', 'netloc', 'path', 'params', 'query', 'fragment']\n"
 '    \n'
 '    def _asdict(self):\n'
 '        return self.__dict__\n'
 '    \n'
 '    @classmethod\n'
 '    def _make(cls, iterable, new=tuple.__new__, len=len):\n'
 '        return new(cls, iterable)\n'
 '    \n'
 '    def _replace(self, scheme=None, netloc=None, path=None, params=None, '
 'query=None, fragment=None):\n'
 '        return self\n'
 '    \n'
 '    def __getnewargs__(self):\n'
 '        return tuple(self)\n'
 "    scheme = property(lambda self: self[0], doc='Alias for field number 0')\n"
 "    netloc = property(lambda self: self[1], doc='Alias for field number 1')\n"
 "    path = property(lambda self: self[2], doc='Alias for field number 2')\n"
 "    params = property(lambda self: self[3], doc='Alias for field number 3')\n"
 "    query = property(lambda self: self[4], doc='Alias for field number 4')\n"
 "    fragment = property(lambda self: self[5], doc='Alias for field number "
 "5')\n")

Therefore to circumvent this false-positive,

1. Check that it is a function call
2. instance's class can be inferred to be derived from tuple
   --> Then we don't fire the error

I'm not too sure if this is the correct approach and thus it's marked as a draft PR, but opening this PR to get reviews / thoughts on whether this might be a potential solution / how to approach this issue

Type of Changes

	Type
✓	🐛 Bug fix

Related Issue

Closes #4377

[coveralls]

Coverage increased (+0.001%) to 91.829% when pulling cac4e5d on yushao2:bug-4377 into 4628b74 on PyCQA:master.

[yushao2]

Marking this as ready for review to get some comments -- to see if this is the correct approach to the issue

[Pierre-Sassoulas]

Look good to me :)

[cdce8p]

I'm not sure this should be fixed in pylint. To me it looks like a issue with the astroid inference of namedtuples. A smiliar one seems to be #4370

@nelfin Do you have an idea how to fix it? If I remember correctly, you've already worked on brain_namedtuple_enum.py.

[nelfin]

I'm not sure this should be fixed in pylint. To me it looks like a issue with the astroid inference of namedtuples. A smiliar one seems to be #4370

@nelfin Do you have an idea how to fix it? If I remember correctly, you've already worked on brain_namedtuple_enum.py.

I'll have a look, it doesn't seem like it's an issue with subclassing NamedTuples at first glance. The sorted_query lines in the test are a red herring as well, I can make the same test give a false-positive with this:

from urllib import parse
parsed_url = parse.urlparse("http://www.this-is-weird.com")
a = parse.ParseResult._replace(parsed_url, query='foo')
b = a.geturl()  # false-positive: no-member

From investigation, it seems like _replace() yields a NamedTuple instance and .geturl() is not a valid attribute -- therefore the warning is triggered.

_replace returns self (in the transformed namedtuple class definition), not a new NamedTuple instance. I think what's happening here is urllib.parse defines a namedtuple and assigns it to a different name (which the astroid namedtuple brain ignore), then later subclasses that base namedtuple. I think then pylint attempts to infer that class definition and it only sees the first one, but I'm less familiar with pylint that I am with astroid.

# snippet from urllib.parse
_ParseResultBase = namedtuple('ParseResult', 'scheme netloc path params query fragment')
# ...
class ParseResult(_ParseResultBase, _NetlocResultMixinStr):
    __slots__ = ()
    def geturl(self):
        return urlunparse(self)

[Pierre-Sassoulas]

Closing because @nelfin convinced me it should be fixed in astroid instead.