python - Why tkinter module raises attribute error when run via command line but not when run via IDLE?

Question

Is there a reason why the code will raise an error when run via the command line compared to when run via IDLE's run module f5 command?

Recently I've been trying to improve the readability and robust-ness of my code. As a result I've been trying to remove all the from module import * lines. I used to use from tkinter import * and this line of my code worked perfectly fine:

self.path = filedialog.askdirectory()

But now I have changed from tkinter import * to import tkinter as tk and I have changed the code accordingly:

self.path = tk.filedialog.askdirectory()

A file called GUI.py imports this file with: from lib.filesearch import * (the line of code I mentioned resides within the filesearch file.)

I run my code via IDLE and everything is fine. My GUI still works and the line self.path = tk.filedialog.askdirectory() works like normal however, when I run the code through windows command line I get the error:

AttributeError: 'module' object has no attribute 'filedialog'

Here are the relevant bits from my code:

From filesearch.py

import tkinter as tk
    def get_path(self):
        """Store user chosen path to search"""
        self.paths = tk.filedialog.askdirectory(initialdir = FileSearch.DEFAULT)
        return self.paths

From GUI.py

from lib.filesearch import *    
    def Browse(self):
        self.BrowseB['state']='disabled'
        self.p=self.CrawlObj.get_path()
        self.AddText('Searching from Path: ' + str(self.p))
        self.BrowseB['state']='normal'

Unlike this question I only have one version of python installed. Namely, Python34.

Answer 1

I want to start by saying: always explicitly import submodules if you know you will use them. The end of this answer has a more compelling case where this is important.

Because of the structure of tkinter you must explicitly import submodules for them to load:

import tkinter as tk
print(hasattr(tk,"filedialog")) # in a standard interpreter will print false
import tkinter.filedialog
print(hasattr(tk,"filedialog")) # should always print true after explicit import

the reason you don't need to do this in IDLE is that before your code is run IDLE sets up some stuff in the background and ends up importing some of the tkinter libraries. One of the maintainers has commented that this is effectively a bug in IDLE.

In python 3.6.5 (and possibly earlier, only checked this version) this specific discrepancy has been fixed so it no longer happens for all but 2 modules I show below.

in any version you can see a list of submodules that are loaded with some code like this:

Python 3.6.5 (v3.6.5:f59c0932b4, Mar 28 2018, 03:03:55)
# standard interpreter 
>>> import sys
>>> len(sys.modules) #total number of modules automatically loaded
71
>>> sorted(name for name in sys.modules.keys() if ("." in name)) #submodules loaded
['collections.abc', 'encodings.aliases', 'encodings.latin_1', 'encodings.utf_8', 'importlib._bootstrap', 'importlib._bootstrap_external', 'importlib.abc', 'importlib.machinery', 'importlib.util', 'os.path']
>>> len(_) #number of submodules
10

And in IDLE:

Python 3.6.5 (v3.6.5:f59c0932b4, Mar 28 2018, 03:03:55) 
# IDLE
>>> import sys
>>> len(sys.modules)
152
>>> sorted(name for name in sys.modules.keys() if ("." in name and "idlelib" not in name))
['collections.abc', 'encodings.aliases', 'encodings.ascii', 'encodings.latin_1', 'encodings.utf_8', 'importlib._bootstrap', 'importlib._bootstrap_external', 'importlib.abc', 'importlib.machinery', 'importlib.util', 'os.path', 'tkinter.constants', 'urllib.parse']
>>> len(_) #number of submodules not directly related to idlelib.
13

tkinter.constants is loaded when you just import tkinter so as of the version I tested, this issue still exists for only urllib.parse and encodings.ascii (and idlelib modules but generally production code doesn't use that)

---

This isn't necessarily an IDLE specific issue though, a worse issue is if the submodule is loaded by another library you use. Take the following code as an example:

>>> import pandas
>>> import http
>>> http.client
<module 'http.client' from '.../http/client.py'>

now lets say we wrote some other code that still used http.client but didn't use pandas:

>>> import http
>>> http.client
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'http' has no attribute 'client'

This way you could end up with a submodule that works properly when the code that uses it loads http.client possibly by using a library that happens to use it but will otherwise fail.

This takes me back to my initial point - always explicitly import submodules.