weird behavior of the import statement

[plzfgme]

Bug report

Bug description:

File structure:

 a
├──  __init__.py
├──  c.py
└──  d.py

File content:

# __init__.py
b = "b"
from .c import c

__all__ = ["b", "c"]

# c.py
c = "c"

# d.py
d = "d"

Run in python:

Python 3.11.9 (main, Jan  1 1980, 00:00:00) [Clang 16.0.6 ] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import a.b as b
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'a.b'
>>> import a.c as c
>>> c
'c'
>>> import a.d as d
>>> d
<module 'a.d' from '/the/path/to/Python-3.11.9/a/d.py'>

I think for b and d, it works correctly because b is not a module and d is. But for c, I expect it to be a module, but it is actually a variable.

I am not familiar to python but I cannot find any info about this behavior, so I ask it here.

CPython versions tested on:

3.11

Operating systems tested on:

Linux

[serhiy-storchaka]

This was discussed recently in #117860. It is not obvious that this is a bug. The code in __init__.py sets c to 'c'. import a.c as c is equivalent to import a.c; c = a.c.

To avoid confusion try to not override the names which correspond to names of submodules with other values.