Importing a python module by path

Question

I have the following mis-fortunate situation:

Three directories (A, B and C) contain a python module M with a function F. (Those directories are not packages and it is impossible to change anything about the situation.)

I am looking for a way to import them separately to access their functionalities. How do I need to import those modules to access F somewhat like this:

A.F()
B.F()
C.F()

• timger: user from a.M import F as A_F
• wim: I'm curious about how you got into this f0rked up situation, why can't you solve it the 'proper' way and now have to hack something together?
• Woltan: @wim The f0rked up situation is a dependency and you do not want to change things in a dependency or you will end up messing up the interface which leads to god knows what... This is why I am looking for a 'proper' way at least on my side of the interface.
• Piotr Dobrogost: Duplicate of How to import a module given the full path?

Answer 1

You need to exec things into a new locals dictionary. You can only get to the files you mention as files, not as modules, then stuff them into a moduletype.

from types import ModuleType
with open("A/M.py") as a:
 A = ModuleType('A')
 exec a.read() in A.__dict__
with open("B/M.py") as b:
 B = ModuleType('B')
 exec b.read() in B.__dict__
with open("C/M.py") as c:
 C = ModuleType('C')
 exec c.read() in C.__dict__

Then access them like B.F() as you wanted. The only problem is the module metadata isn't set up correctly, so it will appear as a builtin. You can even then do:

import sys
sys.modules['A'] = A
sys.modules['B'] = B
sys.modules['C'] = C

and they will be importable like import A from other parts of your application.

• Matthew Wilkes: Added the sys.modules problem as an advantage to this method by explicitly setting the constructed modules in you can reimport.

Answer 2

This will work, but it seems a bit inelegant...

import sys
sys.path.append("A")
import M as A
sys.path.pop()
del sys.modules['M']
sys.path.append("B")
import M as B
and so on...

• mouad: IMO this is the way to go you just have to fix the fact that pre_path point always to sys.path so changing sys.path will change pre_path. What you should do is each time pop() from sys.path after you finish importing.
• Woltan: Somehow the module M in B is not loaded with import M as B if previously import M as A was imported. If I call F of B with B.F() actually the function F in A is called.
• mouad: ahh and you have also to delete sys.modules['M'] after importing it, because the next time you try to import the other module M the import will be skipped because it's already in sys.modules.
• Matthew Wilkes: Probably it's still in sys.modules as 'M'

Answer 3

put an __init__.py in each A/ B/ and C/. The content of this file is

from M import F

Than the following code should work:

import A, B, C
A.F()
B.F()
C.F()

The __init__.py declares the directory as a package, and the statements in this file are executed when you import the package.

• Woltan: Thx for the reply, but as I said, I cannot change anything about how the directories look like. So no __init__.py can be added to make it a package.

Answer 4

use import as like

from A.M import F as A_F

• Woltan: But then the sys.path would have to be set each time I import a module. Could you elaborate on that part?
• timger: @Woltan sorry for my thoughtless you should add your directory to syspath use sys.path.append like the answer from rocksportrocker