Why does a python module act like a singleton?

Question

I create a dictionary from a remote database as part of my application run. This process is pretty I/O heavy, so I've decided to create a "singleton" instance of this dictionary and just call it as it is needed in my application.

The code looks like (in Dictionaries.py):

state_code_dict = None

def get_state_code_dict():
    global state_code_dict
    if state_code_dict == None:
        state_code_dict = generate_state_code_dict()
    return state_code_dict

I then import and call the get_state_code_dict() function where needed. I added a print statement to check if state_code_dict was being reinitialized or reused, and I found it was being reused (which is the functionality I want). Why is the instance of state_code_dict surviving the application run?

Edit

I import the get_state_code_dict function in multiple files.

edited to clarify: I import the code in multiple locations (multiple files include the `get_state_code_dict` function. — Tyler DeWitt, Jun 07 '12 at 17:32
Importing code that the interpreter has already loaded will not re-load that code by default. It's possible to deliberately reload a module but if you have to do this for any reason other than because the actual code of the module might have changed during runtime, you should probably use a different programming idiom to do what you want to do. — Andrew Gorcester, Jun 07 '12 at 17:35
And this behavior is desirable precisely because module initialization is sometimes heavyweight. — Russell Borogove, Jun 07 '12 at 17:40

score 24 · Accepted Answer · edited Jun 21 '16 at 14:33

This is the Python Language Reference's description of how importing a module works:

(1) find a module, and initialize it if necessary; (2) define a name or names in the local namespace

(Emphasis added.) Here, initializing a module means executing its code. This execution is only performed if necessary, i.e. if the module was not previously imported in the current process. Since Python modules are first-class runtime objects, they effectively become singletons, initialized at the time of first import.

Note that this means that there's no need for a get_state_dict_code function; just initialize state_code_dict at top-level:

state_code_dict = generate_state_code_dict()

For a more in-depth explanation, see this talk by Thomas Wouters, esp. the first part — around 04:20 — where he discusses the "everything is runtime" principle.

Would I then get the state_code_dict by calling `import Dictionaries; state_code_dict = Dictionaries.state_code_dict`? — Tyler DeWitt, Jun 07 '12 at 17:57
@TylerDeWitt: yes. Or by the idiomatic `from Dictionaries import state_code_dict`. — Fred Foo, Jun 07 '12 at 18:00

score 14 · Answer 2 · edited Jun 08 '12 at 01:56

I voted larsmans answer, i just wanted to add an example.

hello.py:

hi = 'hello'

print(hi)

def print_hi():
    print(hi)

ipython session:

In [1]: from hello import print_hi
hello

In [2]: print_hi()
hello

In [3]: from hello import print_hi

In [4]: import hello

In [5]: hello.print_hi()
hello

Look that the imports at lines 3 and 4 don't output "hello" as the import in line 1 did, that means the code isn't re-executed.

Why does a python module act like a singleton?

2 Answers2

Linked