Improve the situation for `no-name-in-module` for C extensions and modules using dynamic features

[chrisspen]

I have the ProxyTypes package installed and have a file mymodule.py that imports it like:

from peak.util.proxies import CallbackProxy

However, pylint throws this false error for it:

E:  8, 0: No name 'util' in module 'mymodule.peak' (no-name-in-module)

Pylint should reference the peak package installed into my virtualenv, and not assume it's a local non-existent module.

[PCManticore]

Please provide a minimal working example, I'm not sure I completely understand what is happening, especially since the error you posted indicate something more complex going on than the first import.

[chrisspen]

You can easily reproduce it with the following procedure:

$ mkdir test
$ cd test
$ virtualenv .env
Running virtualenv with interpreter /usr/bin/python2
New python executable in test/.env/bin/python2
Also creating executable in test/.env/bin/python
Installing setuptools, pkg_resources, pip, wheel...done.
$ .env/bin/pip install pylint ProxyTypes
Collecting pylint
  Using cached pylint-1.7.1-py2.py3-none-any.whl
Collecting ProxyTypes
  Downloading ProxyTypes-0.9.zip
Collecting six (from pylint)
  Using cached six-1.10.0-py2.py3-none-any.whl
Collecting mccabe (from pylint)
  Using cached mccabe-0.6.1-py2.py3-none-any.whl
Collecting singledispatch; python_version < "3.4" (from pylint)
  Using cached singledispatch-3.4.0.3-py2.py3-none-any.whl
Collecting isort>=4.2.5 (from pylint)
  Using cached isort-4.2.15-py2.py3-none-any.whl
Collecting astroid>=1.5.1 (from pylint)
  Using cached astroid-1.5.3-py2.py3-none-any.whl
Collecting configparser; python_version == "2.7" (from pylint)
Collecting backports.functools-lru-cache; python_version == "2.7" (from pylint)
  Using cached backports.functools_lru_cache-1.4-py2.py3-none-any.whl
Collecting enum34>=1.1.3; python_version < "3.4" (from astroid>=1.5.1->pylint)
  Using cached enum34-1.1.6-py2-none-any.whl
Collecting lazy-object-proxy (from astroid>=1.5.1->pylint)
  Using cached lazy_object_proxy-1.3.1-cp27-cp27mu-manylinux1_x86_64.whl
Collecting wrapt (from astroid>=1.5.1->pylint)
Building wheels for collected packages: ProxyTypes
  Running setup.py bdist_wheel for ProxyTypes ... done
  Stored in directory: /home/chris/.cache/pip/wheels/e3/8f/4c/5d7b9f8e251cc9567b730e7f9e2f74a9ea36a4d9189e0d2864
Successfully built ProxyTypes
Installing collected packages: six, mccabe, singledispatch, isort, enum34, lazy-object-proxy, wrapt, backports.functools-lru-cache, astroid, configparser, pylint, ProxyTypes
Successfully installed ProxyTypes-0.9 astroid-1.5.3 backports.functools-lru-cache-1.4 configparser-3.5.0 enum34-1.1.6 isort-4.2.15 lazy-object-proxy-1.3.1 mccabe-0.6.1 pylint-1.7.1 singledispatch-3.4.0.3 six-1.10.0 wrapt-1.10.10
$ echo -e "from peak.util.proxies import CallbackProxy\n" > mymodule.py
$ ls -lah
total 16K
drwxrwxr-x  3 chris chris 4.0K Jun 20 16:54 .
drwxr-xr-x 11 chris chris 4.0K Jun 20 16:52 ..
drwxrwxr-x  7 chris chris 4.0K Jun 20 16:53 .env
-rw-rw-r--  1 chris chris   45 Jun 20 16:54 mymodule.py
$ .env/bin/pylint mymodule.py
No config file found, using default configuration
************* Module mymodule
C:  2, 0: Trailing newlines (trailing-newlines)
C:  1, 0: Missing module docstring (missing-docstring)
E:  1, 0: No name 'util' in module 'mymodule.peak' (no-name-in-module)
E:  1, 0: Unable to import 'peak.util.proxies' (import-error)
W:  1, 0: Unused CallbackProxy imported from peak.util.proxies (unused-import)

------------------------------------------------------------------------
Your code has been rated at -120.00/10 (previous run: -120.00/10, +0.00)


$ .env/bin/python
Python 2.7.12 (default, Nov 19 2016, 06:48:10) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from peak.util.proxies import CallbackProxy
>>> 

[meniluca]

I see the same behaviour with another library. Any idea how to fix this?

[andrewxiwu]

Same problem here. Any idea/reference to fix this issue? It is really annoying to have these false positives.

[neocogent]

Same problem here with gevent.ssl.
pylint says SSLContext and PROTOCOL_SSLv23 don't exist but they clearly do.

Python 2.7.12. gevent 1.2.2

python
Python 2.7.12 (default, Nov 19 2016, 06:48:10) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from gevent.ssl import SSLContext, PROTOCOL_SSLv23
>>> 

And another very common one also reported as non-existent, in this case as no-member,

[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import MySQLdb as db
>>> db.Error
<class '_mysql_exceptions.Error'>
>>> 

Have to disable no-name-in-module for now.

Maybe it's related error but I also get a warning on this,
from backports.functools_lru_cache import lru_cache
It says 'relative-import' when it's clearly not.

Incidentally I am not using virtualenv here. This is system wide python.

[soundlake]

It seems duplicated. #1138

[soundlake]

@neocogent --unsafe-load-any-extension=y worked for me.

[rawrgulmuffins]

I'm running into a similar issue at my work with a pure python sub-package. --unsafe-load-any-extension has no effect in this case.

[jcezarms]

I'm an year late, but still seeing the exact same behaviour with tensorflow.python.lib.io, even though both the library and its __init__.py file exist.

As I'm using vscode, this issue causes workspace errors to render in a misleading and incorrect way. To quickly reproduce, install tensorflow and type from tensorflow.python.lib.io import file_io in a newly created file:

Running pylint v1.9.2 from the command line on the above file points the same

$ pylint some.py
No config file found, using default configuration
************* Module some
C:  1, 0: Final newline missing (missing-final-newline)
C:  1, 0: Missing module docstring (missing-docstring)
E:  1, 0: No name 'python' in module 'tensorflow' (no-name-in-module)
W:  1, 0: Unused file_io imported from tensorflow.python.lib.io (unused-import)

-----------------------------------------------------------------------
Your code has been rated at -70.00/10 (previous run: -40.00/10, -30.00)

For these specs

$ pylint --version
No config file found, using default configuration
pylint 1.9.2,
astroid 1.6.5
Python 3.6.6 (v3.6.6:4cf1f54eb7, Jun 27 2018, 03:37:03) [MSC v.1900 64 bit (AMD64)]

@PCManticore I'd consider the above a minimal working example for this issue, can you review it to maybe consider reopening this one?

[brycepg]

• peak issue is due to How should I correctly lint an implicit namespace package? #1361 I think
• gevent.ssl is doing copy_globals(_source, globals()) (wontfix / maybe astroid brain fix)
• tensorflow seems to be a valid issue

[rafaelmendy]

Same problem in importing PyQt5 libraries.

The code works but pylint send errors.