Question

I have a package with a similar structure as scipy/numpy, where you have a few main submodules, and each submodule contains functions from various files that have been flattened into the same namespace. Let’s say

package/
  sub1/
    __init__.py
    file1.py
    file2.py

And then sub1/__init__.py looks like this:

from .file1 import func1, func2
from .file2 import func3

The result is that I can do

import package.sub1
package.sub1.func1()

However, the problem is that the following does not work:

>>> import package.sub1
-- change things in file1.py --
>>> reload(package.sub1)

The function does not update. It works if I do import package.sub1.file1 instead, so there is something with flattening the namespace, that makes it lose the connection to the module. My main question is how I can still use the reload command, while still getting the benefits of being able to write sibling functions in separate files.

Answer 1

Make the __init__.py reload the files it imports when it is imported:

import file1
import file2
reload(file1)
reload(file2)

func1, func2 = file1.func1, file1.func2
func3 = file2.func3

That said, relying on reload() is generally a bad idea if you can avoid it.